Automatic winding device for harness branch adhesive tape

Abstract

The present application relates to the technical field of automatic tape winding control, and discloses a wire harness branch tape automatic winding device, wherein the system simultaneously outputs cross-section geometric parameters to a tension control module and a feeding control module through a profile triangulation assembly, respectively generates a dynamic angular velocity compensation curve and a differentiated feeding speed curve according to square root compensation, and eliminates the non-circular cross-section winding ratio azimuth deviation through multi-loop coupling; a visual detection module simultaneously feeds a defect correction signal to the tension and feeding control modules to realize two-way correction; in the digital twin module, a health monitoring link outputs a health regulation instruction with a wavelet packet energy envelope and a baseline plus 3sigma threshold value, and a process self-learning link trains a GBDT regression model with four types of input features including cross-section, tension, feeding and vision, and reversely issues an optimal process parameter mapping table.

Description

Technical Field

[0001] This invention relates to the field of wire harness tape winding equipment, and more specifically to an automatic wire harness branch tape winding device. Background Technology

[0002] In automotive wiring harnesses, such as slender main wiring harnesses and door wiring harnesses, the surface of the tape is spirally wrapped, which is a common process in automotive wiring harness manufacturing for insulation protection, wear resistance, and moisture protection. Cited patent CN212863654U "includes a frame, a rotating turntable, a geared motor driving the turntable, a wire roller mounting position, and a tensioning assembly. The motor speed is collected by a speed sensor and controlled by a PLC to achieve spiral winding of the tape on the surface of the wiring harness; a guide wheel and tensioning assembly are configured to achieve passive tensioning of the tape." This application adds servo torque active tension closed-loop, laser contour non-circular cross-section dual-channel compensation, rotational vision bidirectional correction, magnetic scale dual-stage pitch closed-loop, and digital twin self-optimization, solving the defects of uneven non-circular cross-section winding, large tension fluctuations, and lack of process quality closed-loop under open-loop control.

[0003] In conventional control methods, tape tension relies on passive mechanical tensioning without active feedback control capabilities. At the beginning and end of the winding process, the tape speed changes drastically, and the passive tensioning mechanism cannot respond in time. Furthermore, there is no real-time online detection mechanism for the entire winding process. The turntable operates at a constant angular velocity, and the winding process lacks a quality closed loop that provides feedback corrections to the execution loop through visual inspection. Axial feed is often set in an open loop with a fixed speed ratio. The rotating mechanism lacks health status monitoring and self-optimization scheduling of process parameters based on historical process data. Each control loop is independent of the others, lacking bidirectional data interaction and multi-loop coupling. Tension fluctuations, uneven winding, pitch deviations, and quality defects are superimposed on each other. Summary of the Invention

[0004] This invention provides an automatic winding device for branch tape of wire harnesses to solve the problem that the winding control effect and process consistency cannot meet the requirements for precision wrapping and winding of multi-specification, non-circular cross-section slender workpieces.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: In a first aspect, an automatic wire harness branch tape winding device includes a main unit housing, a turntable assembly disposed inside the main unit housing, a first servo motor disposed inside the main unit housing and driven to rotate the turntable assembly via a gear transmission mechanism, and a main control unit disposed on the front of the main unit housing. The turntable assembly has a C-shaped wire passage opening on one side and a tape positioning component on one side of the turntable. The device also includes: The tension control module includes a servo tensioning roller assembly for performing closed-loop adjustment of belt tension; The feed control module includes an axial feed mechanism for performing axial feed and pitch control of the wire harness; The cross-sectional geometric parameters are obtained by the contour triangulation measurement component and input to the tension control module and the feed control module. The turntable dynamic angular velocity compensation data and differential feed speed compensation data are output respectively. Based on the turntable dynamic angular velocity compensation data and differential feed speed compensation data, the non-circular cross-section overlapping orientation deviation is eliminated through multi-loop coupling processing. The image of the wrapped tape ring is inspected by a vision inspection module to identify defects and obtain the identification results. The identification results are then input to the tension control module and the feed control module to output defect correction control commands, so as to correct the bidirectional deviation of tension and pitch in real time. Axial displacement data is obtained by a linear displacement sensor and combined with the absolute rotation angle data of the turntable. The data is then input to the feed control module for coupling processing to obtain the actual helical pitch data. Based on the actual helical pitch data, the distance measuring segment sensor constructs a three-dimensional point cloud model of the tape profile after winding, compares it with the preset standard profile to calculate the pitch deviation, and outputs a pitch deviation correction signal to feed control module. Acoustic emission signals and servo drive current signals of the spindle drive mechanism are collected by acoustic emission sensors and servo current sensors, respectively, and input to the spindle health monitoring link of the digital twin module.

[0006] Secondly, a control method for an automatic winding device for wire harness branch tape includes the following steps: S1. The cross-sectional geometric parameters are obtained by the contour triangulation measurement component and simultaneously output to the tension control module and the feed control module; S2. Generate dynamic angular velocity compensation curves and differentiated feed rate curves according to the extreme diameter formula and the square root distribution logic of each half compensation, so that the two compensations can eliminate the non-circular cross section overlay ratio azimuth deviation after multi-loop coupling. S3. The tension control module actively adjusts the tape tension in a closed loop based on the feedback from the servo torque sensor, and drives the turntable assembly to rotate and wind according to the dynamic angular velocity compensation curve. S4. The visual inspection module performs image acquisition and defect identification, and simultaneously outputs defect correction signals to the tension control module and the feed control module. S5. The feed control module couples the axial displacement data of the magnetic scale with the absolute rotation angle data of the turntable to form a closed loop of the pitch inner loop, and receives the outer loop verification value of the distance measuring segment sensor for compensation. S6. The digital twin module aggregates process data, outputs health control instructions and optimal process parameter mapping table, and sends them back in reverse.

[0007] The above-described solution of the present invention has at least the following beneficial effects: The servo torque sensor and the servo tension roller form an active closed-loop tension control module, which improves the tension control effect of the tape, effectively eliminates tension fluctuations during the acceleration and deceleration stages at the beginning and end of the winding process, completely eliminates quality defects such as tape slack, bulging and tearing, and improves the consistency of winding throughout the entire process.

[0008] The visual inspection module moves the quality inspection node from post-inspection to the winding process, and identifies four types of defects in real time: gaps, overlapping deviations, edge offsets, and wrinkles. The inspection results are immediately fed back to the tension control and feed control to perform the next turn correction. This achieves a closed loop of defect detection in the current turn, correction in the next turn, and verification after correction, minimizing the defect outflow rate and significantly reducing post-inspection and rework costs.

[0009] The laser profile sensor enables online sensing of the geometric parameters of the wire harness cross section. Combined with the dynamic rotation speed compensation curve algorithm and differentiated feed speed control, it collaboratively eliminates the azimuth deviation of the non-circular cross section overlay ratio from two dimensions: rotation speed and feed, achieving uniform and consistent full circumferential coverage of various cross section wire harnesses such as flat bundles and elliptical bundles.

[0010] The combination of magnetic scale and absolute encoder coupling inner loop closed loop and laser ranging outer loop online verification improves the helical pitch control effect, supports seamless online switching of three modes: interval winding, half-overlap winding, and full-overlap winding, fully meets the precision winding process requirements of different wire harness specifications, and shortens the changeover and debugging time by combining self-learning optimization.

[0011] The multi-sensor fusion health monitoring system enables online early warning of damage to spindle bearings and transmission components, transforming passive failure shutdowns into proactive preventive maintenance, reducing the risk of sudden shutdowns, extending equipment lifespan, and reducing maintenance costs and spare parts inventory pressure.

[0012] Through real-time data interaction via the control bus, the modules can perceive, complement, and verify each other. This enables real-time correction of tension and cross-sectional data to simultaneously drive the feed. Visual measurement provides double protection for the magnetic scale, and continuous self-optimization is driven by all process data. Health prediction, in turn, ensures the reliability of the detection. The overall control effect and quality stability of the system are better than the sum of the effects of each module acting alone, demonstrating the added value of system-level collaborative innovation.

[0013] By utilizing a contour triangulation measurement component, the cross-sectional geometric parameters are simultaneously output to the tension control module and the feed control module. Each of the two compensation paths bears half of the compensation amount, and the non-circular cross-section overlay ratio orientation deviation is eliminated through multi-loop coupling. The vision inspection module simultaneously feeds back the defect correction signal to the tension control module and the feed control module, realizing bidirectional real-time correction of tension and pitch by visual quality information. The digital twin module gathers all process data and outputs process optimization parameters and health control commands in reverse. Each loop forms bidirectional data interaction and multi-loop coupling through the control bus, improving the overall control effect and process consistency. Attached Figure Description

[0014] Figure 1 This is a perspective view of the automatic wire harness tape winding device provided in this embodiment; Figure 2 This is a three-dimensional view of the turntable assembly provided in this embodiment; Figure 3 This is a perspective view of the support plate provided in this embodiment; Figure 4 This is an exploded perspective view of the turntable assembly and support plate combination provided in this embodiment; Figure 5 This is a perspective view of the tabletop provided in this embodiment; Figure 6 This is a perspective view of the axial feed mechanism provided in this embodiment; Figure 7 This is an exploded perspective view of the axial feed mechanism provided in this embodiment; Figure 8 This is a flowchart of the control method provided in this embodiment; Figure 9 This is a block diagram of the control system provided in this embodiment; In the diagram: 1. Main unit housing; 2. Main controller; 3. Emergency stop knob; 4. Switch; 5. Protective cover; 6. Turntable assembly; 7. Support platform; 8. Axial feed mechanism; 9. Motor; 10. Lead screw; 11. Wiring harness clip; 12. Magnetic scale; 13. Turntable; 14. Table panel; 15. Cutter; 16. Wireless image transmission module; 17. Miniature industrial camera; 18. Distance measuring joint sensor; 19. First servo motor; 20. Main gear; 21. Slave gear; 22. Support shaft; 23. C-type gear disc; 24. Guide cover; 25. Support plate; 26. Belt mounting shaft; 27. Reducer; 28. Miniature servo motor; 29. ​​Servo torque sensor; 30. Tension roller; 31. Rubber sleeve; 32. Battery pack; 33. Belt threading port; 34. Laser profile sensor; 35. Acoustic emission sensor. Detailed Implementation

[0015] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0016] like Figures 1-9 As shown, the present invention provides an automatic wire harness branch tape winding device, including a main unit housing 1 and a main control unit 2 disposed on the front of the main unit housing 1. A turntable assembly 6 is disposed inside the main unit housing 1. A first servo motor 19 is disposed inside the main unit housing 1, and the first servo motor 19 drives the turntable 13 inside the turntable assembly 6 to rotate through a transmission component. A tape winding component is disposed on one side of the turntable 13 to follow the rotation of the turntable assembly 6 to wind the tape. A C-shaped wire passage opening is provided on one side of the turntable assembly 6 for the wire to be wound. The wire harness passes through, placing it at the rotation center of the winding mechanism to ensure that the tape can completely wrap around the circumference of the wire harness. An emergency stop knob 3 and a switch 4 are provided on one inclined side of the main unit housing 1. A protective cover 5 is flipped on the main unit housing 1. The drive shaft of the first servo motor 19 is connected to a main gear 20. Two support shafts 22 are connected to the inner wall of the main unit housing 1. A driven gear 21 is rotatably provided on the outer side of the two support shafts 22. The driven gear 21 meshes with the main gear 20 and also meshes with a C-shaped gear disk 23.

[0017] Specifically, the main unit housing 1 can provide stable support and protection for the first servo motor 19. The first servo motor 19 can drive the main gear 20 to rotate forward using the drive end. The main gear 20 can drive the driven gear 21 to rotate in the opposite direction under the support of the support shaft 22 using rotational power and meshing action. The two driven gears 21 can work together to continuously drive the C-shaped gear disk 23 to rotate using rotational power and meshing action, so that the C-shaped gear disk 23 can drive the turntable 13 to rotate under the rotational support of the turntable assembly 6. In an emergency, staff can rotate the emergency stop knob 3 to stop the operation of the first servo motor 19, while switch 4 can start the automatic tape winding device.

[0018] The tape positioning and guiding component includes a guide cover 24 disposed on one side of the turntable 13. The guide cover 24 can rotate with the turntable 13. A tape-passing opening 33 is provided through the guide cover 24 for the tape to pass through. A support plate 25 is fixed on the guide cover 24. A tape mounting shaft 26 and a tension roller 30 are rotatably disposed on one side of the support plate 25. The tension roller 30 is located between the tape mounting shaft 26 and the guide cover 24. A rubber sleeve 31 is disposed on the outer side of the tension roller 30. A servo torque sensor 29 is embedded in the support plate 25 and is located on the outer side of the tension roller 30. A reducer 27 is fixed on one side of the support plate 25 and is axially connected to the tension roller 30. A micro servo motor 28 is connected to the input end of the reducer 27 to drive the tension roller 30 to rotate.

[0019] Specifically, the turntable 13 provides stable support for the guide cover 24, the guide cover 24 provides space for the tape threading opening 33, the tape threading opening 33 provides a through channel and guide for the tape, the support plate 25 provides support for the tape mounting shaft 26 under the support of the guide cover 24, the tape mounting shaft 26 provides space for the tape to be fixedly assembled, the tension roller 30 provides support and tension for the tape under the support plate 25, and the turntable 13 can drive the support plate 25, the tape mounting shaft 26 and the tension roller 30 to rotate together through the guide cover 24, utilizing the rotation... The tape is wound onto the wire harness. The servo torque sensor 29 can measure the tension force acting on the tape in real time (equivalent to the shaft end torque divided by the roller radius). The micro servo motor 28 and the reducer 27 can rotate together with the support plate 25. The micro servo motor 28 can drive the tension roller 30 to rotate under the support of the support plate 25 through the reducer 27. The width of the tension roller is not less than the width of the tape. The tape tension is precisely adjusted by controlling the output torque of the motor (i.e., controlling the resistance torque applied to the tape by the tension roller through the rubber sleeve 31). The tension adjustment range is 0 to 20 N.

[0020] The turntable assembly 6 includes a turntable 13 rotatably disposed on its inner side. A table panel 14 is provided on the inner side of the turntable 13 extending to one side. A cutter 15 is provided on one edge of the table panel 14. A C-shaped toothed disc 23 is provided on the side of the turntable 13 near the support plate 25.

[0021] Specifically, the turntable assembly 6 provides rotational support for the turntable 13, the turntable 13 provides stable support for the tabletop 14, the tabletop 14 can rotate with the turntable 13 and provides stable support for the cutter 15, the cutter 15 can facilitate workers to cut the wrapped tape, and the C-shaped toothed disc 23 can use rotational power to drive the turntable 13 to rotate under the rotational support of the fixed side of the turntable assembly 6.

[0022] The visual inspection module includes a wireless image transmission module 16 embedded on one side of the table panel 14, a miniature industrial camera 17 embedded on the table panel 14 by a fixing component, a laser rangefinder verification sensor 18 for the rangefinder joint sensor, a laser rangefinder verification sensor 18 set on the other side of the table panel 14, and a battery pack 32 replaceably set on one side of the support plate 25.

[0023] Specifically, the wireless image transmission module 16 (5.8 GHz Wi-Fi 6) transmits the image stream concurrently to the main control unit 2 via an independent high-priority UDP data channel. The sampling data frame rate is no less than 1 kHz, and the wireless transmission delay is no more than 5ms. No additional wireless hardware is required. The miniature industrial camera 17 can acquire images of each winding state in real time with the support of the table panel 14, and identify four types of defects online: gaps, overlapping deviations, edge offsets, and wrinkles. This enables closed-loop detection and real-time correction of winding quality on a winding-by-winding basis, moving defect detection from post-inspection to the winding process. The laser ranging verification sensor 18 performs non-contact ranging (laser triangulation principle) at the edge of the spiral of the wound tape in the completed area. The range is 10-100 mm, and the measurement accuracy is better than ±0.1 mm. This verifies the actual spiral pitch online and transmits the data to the main control unit 2. The battery pack 32 outputs 5V, 12V, and 24V as required by each component through the onboard DC / DC step-up and step-down voltage regulator module. V DC power supply; the battery compartment adopts a quick-release buckle design, which can complete the removal and replacement of the battery pack 32 without the use of tools. The battery pack 32 can work continuously for no less than 4 hours under rated comprehensive load; the rotating side power supply is completely self-sufficient. The entire power supply scheme has no conductive slip rings and no electrical connections that cross the rotating interface, completely eliminating the wear and maintenance problems of the rotating electrical interface.

[0024] A support platform 7 is provided on one side of the main unit housing 1. An axial feed mechanism 8 is fixed on the support platform 7. The axial feed mechanism 8 includes a motor 9 provided on one side. A lead screw 10 is rotatably installed inside the axial feed mechanism 8 and is axially connected to the output end of the motor 9. A wire harness clip 11 is threaded to the outside of the lead screw 10. A slide is provided on the front of the axial feed mechanism 8. The wire harness clip 11 is guided through the slide. A magnetic scale 12 is provided inside the axial feed mechanism 8. The reading head of the magnetic scale 12 is connected to the wire harness clip 11. The magnetic scale of the magnetic scale 12 is fixed inside the slide and cooperates with the reading head.

[0025] Specifically, the support platform 7 provides stable support for the axial feed mechanism 8 under the support of the main unit housing 1. The axial feed mechanism 8 provides stable support for the motor 9 and provides space for the slide rail. The slide rail provides movement space and guide limit function for the wire harness card 11. The axial feed mechanism 8 provides stable support for the magnetic scale 12. After starting, the motor 9 drives the lead screw 10 to rotate using the drive end. The lead screw 10 drives the wire harness card 11 to move axially according to the rotation direction by means of the thread action with the wire harness card 11. This allows the wire harness card 11 to drive the reading head to move along the magnetic scale 12, so that the reading head can detect the movement. The wire harness card 11 provides a through channel for the wire harness and provides a blocking function for the connector at the end of the wire harness, thus blocking the wire harness and facilitating the movement of the wire harness.

[0026] The contour triangulation assembly includes three laser contour sensors 34 evenly arranged on the fixed side of the turntable assembly 6 to perform multi-directional contour scanning of the wire harness cross section from three directions to detect the wire harness contour.

[0027] An acoustic emission sensor 35 is installed on the fixed side of the turntable assembly 6. The acoustic emission sensor 35 and the servo current sensor can detect the health status of the spindle and transmission components in real time, so as to realize preventive maintenance and early warning. At the same time, through a multi-dimensional process data knowledge base and machine learning algorithms, the process parameters are continuously driven to optimize themselves, shorten the changeover and debugging time, and improve the consistency of process quality between batches.

[0028] The main control unit 2 has a built-in tension control module, feed control module and digital twin module, and is connected to the following sensors and actuators through a control bus to form a multi-sensor closed-loop collaborative control system.

[0029] In one embodiment of the present invention, a servo tensioning roller assembly (the hardware entity of the tension control module) is installed on the support plate 25 to replace the original passive tensioning guide roller assembly. The servo tensioning roller assembly is driven by a micro servo motor 28 mounted on the support plate 25, which drives the tensioning roller 30 to rotate actively via a reducer 27. The roller of the tensioning roller 30 is covered with a rubber sleeve 31. The reducer 27 is fixed to one side of the support plate 25, and its output shaft is axially connected to the tensioning roller 30, while its input end is connected to the micro servo motor 28. By controlling the output torque of the micro servo motor 28, the resistance torque applied to the conveyor belt by the tensioning roller 30 is adjusted, thereby achieving active control of the conveyor belt tension. A servo torque sensor 29 is embedded in the support plate 25 and located on the outside of the tensioning roller 30. It collects the shaft end torque signal in real time, converts it into a conveyor belt tension feedback value, and transmits it to the tension control module in the main controller via a wireless channel, forming an active tension closed loop.

[0030] In one embodiment of the present invention, a contour triangulation measurement assembly is installed on the slender workpiece entry side directly in front of the turntable assembly 6, including three laser contour sensors 34. The three laser contour sensors 34 respectively perform contour scanning of the slender workpiece cross-section in a diameter direction. The three contour scanning data are then processed into point cloud fitting to obtain the slender workpiece cross-section type and major axis dimension. a minor axis dimensions b The cross-sectional profile equation is also calculated. These cross-sectional geometric parameters are simultaneously output to both the tension control module and the feed control module.

[0031] Specifically, taking a slender workpiece with an elliptical cross-section as an example, let the radius of the major axis of the cross-section be... a The minor axis radius is b The rotation angle of turntable 13 is θ (Measured from the major axis), then the cross section is at θ The polar radius at the corner is: Take the equivalent reference polar diameter The physical conditions required for uniform overlay ratio are... .

[0032] This invention employs a dual-channel coupling scheme where rotational speed compensation and feed rate compensation each bear half of the compensation amount, distributed according to the square root: The dynamic angular velocity compensation curve generated by the tension control module is as follows: in This is the reference angular velocity.

[0033] The feed control module synchronously generates differentiated feed rate curves, which drive the axial feed mechanism to execute the corresponding speeds. in This is the baseline feed rate.

[0034] Therefore, in the polar radius is The azimuth and angular velocity compensation coefficient are taken as follows: The feed rate compensation coefficient is taken as The two compensation paths, after being coupled through multiple loops, jointly satisfy the requirements. .

[0035] Furthermore, in the long axis orientation θ =0° ρ ( θ )= a At maximum, angular velocity compensation slows down turntable 13, while feed rate compensation speeds up the feed; in the minor axis orientation... θ =90° ρ ( θ )= bAt the minimum, angular velocity compensation accelerates turntable 13, while feed rate compensation slows down the feed. The tension control module operates as described above. ω ( θ The drive command for the first servo motor 19 is dynamically adjusted angle by angle; the feed control module performs the above-mentioned... Differentiated speed adjustment is performed on the axial feed mechanism 8.

[0036] The tension control module also compares the measured value of the belt tension obtained by the servo torque sensor 29 with the tension set value, and drives the servo tensioning roller assembly to actively adjust the belt tension in a closed loop. The tension closed loop control accuracy is not less than ±0.1N, and the sampling period of the main controller is not greater than 1ms.

[0037] In one embodiment of the present invention, the visual inspection module includes a miniature industrial camera 17 fixed on the inner side table panel 14 of the turntable 13. The miniature industrial camera 17 has a ring-shaped LED lighting module fixed on its upper end, a wireless image transmission module 16 embedded on one side of the table panel 14, a wireless receiving module installed on the fixed side of the main unit housing 1, and a battery pack 32 that can be replaced and disposed on one side of the support plate 25.

[0038] The miniature industrial camera 17 and the ring-shaped LED lighting module rotate synchronously with the turntable 13, acquiring images of the tape wrapping area in real time. The wireless image transmission module 16 transmits the image data acquired by the miniature industrial camera 17 and the real-time sampling data from the servo torque sensor 29 to the wireless receiving module in a concurrent wireless data stream manner. The battery pack 32 rotates synchronously with the support plate 25 and is fixed with quick-release buckles. Its output is distributed into multiple DC power supplies by the DC / DC voltage regulator module, providing unified power to the miniature industrial camera 17, the ring-shaped LED lighting module, the wireless image transmission module 16, and the servo tensioning roller assembly of the tension control module.

[0039] Specifically, the image processing and defect recognition unit is integrated into the main control unit. It receives the image data stream forwarded by the wireless receiving module and performs online defect recognition on each frame of the image. The recognition objects include: gap defects with uncovered bare areas between adjacent tape loops, overlapping deviation defects where the actual overlapping width deviates from the target value, edge offset defects where the tape edge exceeds the allowable axial deviation zone, and wrinkle defects with irregular fold patterns on the tape surface.

[0040] Upon detecting a defect, the vision inspection module simultaneously outputs defect correction signals to both the tension control module and the feed control module: when wrinkles or gaps related to tension are detected, the tension deviation correction amount is transmitted to the tension control module to adjust the target tension value; when overlapping deviation defects are detected, the actual pitch of the current rotation is calculated synchronously, the pitch deviation amount is transmitted to the feed control module, and added to the error integral term of the inner loop of the magnetic scale 12 to compensate for the feed speed of the next rotation, forming a double-insurance mechanism for pitch accuracy involving vision measurement, the magnetic scale 12 closed loop, and feed execution. This achieves bidirectional real-time correction of tension and pitch using visual quality information.

[0041] In one embodiment of the present invention, the magnetic scale 12 is mounted on the axial feed mechanism 8. The magnetic scale is fixed along the inner side of the slide rail. The reading head is connected to the wiring harness 11 and moves linearly with the wiring harness 11, transmitting the axial displacement signal of the slender workpiece to the feed control module in real time. The precise absolute rotation angle of the turntable 13 is obtained by reading the position value of the multi-turn absolute encoder built into the drive shaft end of the first servo motor 19 and converting it according to the transmission ratio of the gear transmission mechanism (the ratio of the number of teeth of the main gear 20 to the number of teeth of the C-type gear disk 23). The feed control module couples the displacement signal of the magnetic scale 12 with the absolute rotation angle signal of the turntable 13 to calculate the actual helical pitch. Where Δ d For turntable 13 to rotate Δ The actual axial feed rate of the slender workpiece during the angle period. With the target pitch as the control target, the drive motor 9 of the axial feed mechanism 8 is controlled in a closed loop to form an inner closed loop of the pitch.

[0042] Specifically, the distance measuring sensor 18 is located on the other side of the platform 14, rotating synchronously with the turntable 13 and aligning with the edge of the wound tape spiral. It independently verifies the actual pitch online, and the verification result is input as the outer loop compensation amount into the feed control module, forming a dual-stage pitch precision control system. When the deviation between the verified pitch and the target pitch exceeds a set threshold (e.g., 0.3mm), the outer loop compensation amount is superimposed. The axial feed mechanism of the feed control module supports seamless online switching between three pitch modes: intermittent winding, half-overlap winding, and full-overlap winding.

[0043] In one embodiment of the present invention, the digital twin module includes an acoustic emission sensor 35, a servo current sensor, a digital twin engine, and a self-learning optimization module. The acoustic emission sensor 35 is mounted on the inner surface of the housing near the spindle bearing on the fixed side of the main unit housing 1; the servo current sensor is mounted on the output side of the first servo motor 19 driver. Two types of implementable data processing links run within the digital twin module: one is a signal processing and threshold decision link for spindle health monitoring; the other is a gradient boosting decision tree (GBDT) regression modeling link for self-optimization of tape spiral winding process parameters; optionally, a lightweight convolutional neural network (CNN) adaptation link for defect trend prediction can also be run. The following fully discloses the input data, processing logic, output results, input-output causal relationship, and technical effects.

[0044] 6.1 Spindle Health Monitoring Link Input data: ultrasonic time-domain waveform output by acoustic emission sensor 35, with a sampling rate of not less than 1MHz, an effective analysis frequency band of 50-500kHz, and a resonant center of approximately 150kHz; three-phase current timing output by servo current sensor, with a sampling rate of not less than 10kHz.

[0045] Processing logic: The digital twin engine performs wavelet packet time-frequency decomposition on the acoustic emission signal. The mother wavelet uses Daubechies-4, with 4 decomposition levels, and extracts the energy envelope feature values ​​of each sub-band in the range of 50-500kHz. ( i =1,2,…, N , N (Number of sub-bands). A statistical baseline mean of the energy envelope for each frequency band is established based on data from at least 100 hours of continuous operation during the initial health phase of the equipment. with standard deviation Dynamic alarm threshold At the same time, the three-phase current is converted into an estimated value of the motor load torque. ( The torque constant of the motor. (Equivalent value of quadrature-axis current), and the statistical average of torque under normal operating conditions. As a baseline for torque monitoring.

[0046] Output result: When any sub-band satisfies At that time, the output bearing early damage health regulation command controls the turntable assembly 6 to reduce its speed to 60% of its rated speed and triggers a preventive maintenance warning; when Continue to exceed If the duration exceeds 3 seconds, output an abnormal load health control command and record a timestamped event log.

[0047] Input-output causal relationship and technical effect: Early pitting corrosion of bearing rolling elements and raceways will cause a sudden increase in acoustic emission energy in the ultrasonic frequency band, therefore... There is a direct physical causal relationship between exceeding the threshold and early damage; sudden load changes or mechanical jamming can raise [the risk of injury]. I q Thus raising Therefore, there is a direct electromechanical causal relationship between excessive torque and abnormal load. This link transforms post-fault shutdown into preventative scheduling with verifiable thresholds, providing healthy boundary constraints for the stable operation of the tension control module and the feed control module.

[0048] Process self-learning GBDT regression modeling link This documentation focuses on improvements in model building and training, fully documenting input features, model structure, training steps, hyperparameters, output mapping, and causal relationships.

[0049] Specifically, the input data settings (feature engineering) are combined with the tape spiral winding control scenario: Historical batches are archived using workpiece specifications as the primary key. The primary key fields are: cross-section type (circular, elliptical, flat, one-hot encoded) and major axis. a short axis b Major-minor axis ratio a / b Pitch mode (interval winding, half-overlap winding, full-overlap winding, one-hot encoding), target pitch .

[0050] The input feature vector x for each training sample includes: Cross-sectional geometry sub-vectors: one-hot encoding of cross-sectional type a、b、a / b Polar diameter sequence Mean, variance, and maximum and minimum radius ratio of uniformly sampled points within a circle (e.g., 360 points, angular resolution 1°); Tension process sub-vector: target tension value Tension overshoot, steady-state tension deviation standard deviation, number of first convergence cycles, and statistics (mean and peak value) of the PID output sequence. Feed process sub-vector: Reference feed rate Statistical analysis of differential feed compensation sequence, mean and standard deviation of the distance error within the 12-inch magnetic scale, and number of laser ranging outer ring correction triggers; Visual quality sub-vectors: the number of occurrences, maximum deviation, and percentage of defect circles for four types of defects in this batch: gap, overlap deviation, edge offset, and wrinkle.

[0051] Corresponding supervision labels (target output vector) y The optimal combination of process parameters, including the optimal target tension value, is the one that achieves the best overall quality for this batch. Optimal reference angular velocity Optimal reference feed rate And the discrete point set of the cross-section-velocity mapping table under the square root distribution of each half-compensation. Comprehensive quality score. Q Defined as: in This represents the percentage of defective cycles. This represents the average pitch error. For the steady-state standard deviation of tension, As weight and Preferred During training, select the parameters with the highest Q from historical batches of the same specification as positive sample labels.

[0052] Model structure (necessary modules and hierarchy): The Gradient Boosting Decision Tree (GBDT) regressor is used as the main model, and... M The addition of regression trees is as follows: Where y0 is the initial value of the label mean. For the first m CART regression tree, This represents the shrinkage learning rate coefficient for the tree. The maximum depth of each tree. Optimal choice 5, minimum number of samples per leaf node During feature splitting, the optimal split point is selected based on the reduction in squared error. Number of trees M =100 ~ 200, preferably 150. This structure nonlinearly maps high-dimensional process features into directly deployable process parameters, avoiding the risk of overfitting deep neural networks on small sample industrial data.

[0053] Training process and core parameters: Sample triggering conditions: When the cumulative number of valid batches under the same workpiece specification primary key is no less than 5, a retraining is triggered once; thereafter, incremental retraining is performed for every additional 3 batches. Data partitioning: 80% of the samples are used as the training set and 20% as the validation set in chronological order. Random shuffling is prohibited to avoid information leakage in future batches. Loss function: Multi-output weighted mean square error ,in k Traversal Equal output components, After normalization by dimensions, assign equal weights or allocate according to on-site importance; Optimization method: Forward stepwise gradient boosting, fitting the current residual in each round; learning rate The global value is 0.05. 0.1, preferably 0.08; early stopping is adopted: if the validation set loss does not decrease for 10 consecutive rounds, the process stops; Regularization and overfitting prevention: subsampling ratio 0.8, column sampling ratio 0.8; for high-dimensional statistics of extreme diameters, principal component compression is performed to reduce them to no more than 16 dimensions before inputting them into the model.

[0054] Output result: The model outputs an optimal process parameter mapping table for this specification, which includes at least: and according to and Discretized angular velocity and feed feedforward curve templates. After the mapping table is verified by the main controller for range validity, it is sent back to the tension control module and the feed control module as the initial control parameters (feedforward) for subsequent batches of the same specifications. The PID is only fine-tuned within a small range in the feedforward neighborhood.

[0055] Input-output causal relationship and technical effect: The cross-sectional geometry and extreme diameter statistics in the input determine the non-circular compensation feedforward shape; tension and feed process statistics reflect the convergence of this specification under the existing control law; visual defect statistics directly measure the winding quality. Output parameters shorten the number of convergence loops in the first turn and reduce [the number of loops] through feedforward. and This will improve the overall quality score. Q The causal chain is as follows: historical batches (sections, processes, and quality), GBDT fitting of the optimal parameter surface, distribution of feedforward, reduction of online trial and error, and improved consistency between batches. Those skilled in the art can reproduce the training and deployment without additional creative effort based on the above feature definitions, tree depth, number of trees, learning rate, loss, and early stopping rules.

[0056] One embodiment of the present invention, a defect trend prediction CNN adaptation link This link is an adaptation and improvement of existing artificial intelligence models in the scenario of tape spiral winding control. The input and output settings, scenario adaptation logic, output, and internal connections are disclosed.

[0057] Input / output data settings: The input is an image block (e.g., a 128×128 pixel grayscale image) uploaded by the vision inspection module, covering the tape edge and the gap area between adjacent loops, and stitched together with a snapshot of the current loop's process parameters. The defect history one-hot encoding is embedded into a conditional vector through a fully connected layer; the output is the probability of occurrence of each of the four types of defects in the next round. ( j = Gap, overlap deviation, edge offset, wrinkles) and suggested intervention type (increase tension, decrease tension, or compensate feed).

[0058] Algorithm adaptation logic tailored to specific scenarios: The backbone network uses a pre-trained lightweight CNN (e.g., the first few depthwise separable convolutional blocks of MobileNetV2), with the input channel changed to 1 to adapt to grayscale images. After global average pooling, a fully connected head with embedded process parameters is connected. The hidden layer width is 128, the activation function is ReLU, and the output layer is Softmax / Sigmoid. Training employs transfer learning: the backbone is first frozen, and only the fully connected head is trained for a few epochs (e.g., 10 epochs), then a smaller learning rate (e.g., ...) is applied. Unfreeze the last two convolutional blocks and fine-tune them together. The loss function is a weighted binary cross-entropy, with increased weights for rare defect classes. It is recommended to start effective prediction after accumulating at least 500 images labeled with defect types and process parameters.

[0059] Output Results and Interventions: When a certain type When the threshold is exceeded (e.g., 0.6), an intervention command is sent to the tension control module and / or feed control module one revolution in advance. This is isomorphic to the defect correction signal, thus intervening in the process of predicting the impending defect.

[0060] The inherent relationship between input and output: Image features such as the gradual reduction of grayscale gradient at the edge of the tape and the widening of the gap area are causally related to insufficient tension and excessive pitch. By jointly inputting process parameter snapshots and image features, the model learns under what tension-pitch conditions what texture trends will evolve into defects, thereby establishing an interpretable relationship between image representation and control quantity, rather than treating CNN as an unrepeatable black box labeler.

[0061] In one embodiment of the present invention, the tension control module, feed control module, vision inspection module, and digital twin module form bidirectional data interaction and multi-loop coupling through a control bus, specifically including: Contour triangulation measurement component, tension control module and feed control module: The cross-sectional geometric parameters are simultaneously output in dual channels, driving angular velocity compensation and feed differential compensation respectively; The visual inspection module, tension control module, and feed control module simultaneously output defect correction signals in dual channels to adjust the target tension value and the feed speed for the next revolution, respectively. Tension control module, feed control module, vision inspection module, digital twin module: all process data converge in a positive direction; Digital twin module, tension control module and feed control module: reverse transmission of process optimization parameters; digital twin module and turntable assembly 6: preventive scheduling of health control commands.

[0062] The aforementioned links enable tension regulation, speed compensation, pitch control, visual correction, and health scheduling to operate as an inseparable, collaborative whole.

[0063] In one embodiment of the present invention, the control method of the automatic winding device for wire harness branch tape based on the above control is implemented according to the following steps: S1, Cross-section geometry sensing: The contour triangulation measurement component performs multi-directional scanning on the slender workpiece to be wound to obtain the major axis dimension of the cross-section. a minor axis dimensions b The cross-sectional profile equation is calculated, and the cross-sectional geometric parameters are simultaneously output to the tension control module and the feed control module. S2, Generation of dual-channel compensation for non-circular cross-sections: based on the extreme diameter formula. Calculate each angle θ The extreme diameter of the cross section at that location is taken as follows: Tension control module according to Generate dynamic angular velocity compensation curves; feed control module according to Generate differentiated feed rate curves; speed compensation and feed rate compensation each bear half of the compensation amount, and are satisfied through multi-loop coupling. ; S3, Active Closed-Loop Tension and Rotary Winding: The tension control module compares the measured value of the tape tension obtained by the servo torque sensor with the tension set value, drives the servo tension roller assembly to actively adjust the tape tension in a closed loop, and drives the turntable assembly to rotate and wind according to the dynamic angular velocity compensation curve. S4, Rotational Vision Bidirectional Correction: The vision inspection module performs real-time image acquisition and defect identification on the winding tape loop, outputs tension-related defect correction signals to the tension control module to adjust the tension target value, outputs pitch-related defect correction signals to the feed control module to compensate for the feed speed of the next loop; when an overlap deviation defect is identified, the actual pitch of this loop is calculated synchronously and the pitch deviation is added to the inner ring error integral term of the magnetic scale of the feed control module; S5, Pitch Dual-Stage Closed Loop: The feed control module couples the axial displacement signal of the magnetic scale linear displacement sensor with the absolute rotation angle signal of the turntable to obtain the actual helical pitch and form an inner closed loop. At the same time, it receives the outer loop verification value from the distance measuring segment sensor to compensate for the feed speed target value. S6, Digital Twin Reverse Scheduling: The digital twin module gathers all process data from steps S1 to S5, outputs health control instructions according to the health monitoring link in Section 6.1, and feeds the input features x, composed of cross-sectional geometry, tension process, feed process, and visual quality sub-vectors, into the trained GBDT regression model according to Section 6.2, outputs the optimal process parameter mapping table, and sends it back to the tension control module and feed control module; when the accumulated quality image samples reach the activation condition, the defect trend prediction in Section 6.3 can be executed and intervention instructions can be issued in advance.

[0064] Slender workpiece with elliptical cross section a=18.2mm b Taking 11.8mm as an example, Distribute according to the square root of each half: in the direction of the major axis , The rotational speed dropped to approximately 89.8%. The feed rate increased by approximately 11.4% in the minor axis orientation. The rotational speed increased by approximately The feed rate is reduced by approximately 10.3%. After the two paths are coupled, the overlay ratio tends to be consistent in all directions.

[0065] Comparative Examples, Implementation Examples and Comparative Analysis To verify the advantages of this invention over the existing control system described in the background art, and to specifically illustrate how the improvement in control precision and data processing accuracy is manifested, the following comparative examples and embodiments are provided. Unless otherwise specified, the tape specifications (width 19mm, thickness 0.13mm, PVC adhesive backing) and workpiece specifications (door panel cable, elliptical cross-section) are consistent. Target pitch mode (semi-lapped winding) ), target tension Rotary table reference speed The effective number of winding turns per piece was 48, and the ambient temperature was kept constant at (23±2)℃. Ten pieces were repeated in each group, and the arithmetic mean was taken.

[0066] Evaluation index definition Tension control accuracy The entire winding process is sampled at 1kHz, which satisfies... The percentage of sampling points; and simultaneously, the steady-state standard deviation of tension. Maximum overshoot during start and end segments .

[0067] Pitch control accuracy : Statistical by circle, satisfying The percentage of laps; and the average absolute error of pitch. .

[0068] Overlap ratio azimuth range Divide the circle into 24 azimuths at 15° intervals and calculate the overlay ratio for each azimuth. ,Pick

[0069] Defect outflow rate Number of defective parts confirmed by offline re-inspection after leaving the workstation / total number of parts; percentage of process defect cycles. This represents the number of defective turns detected during the winding process / the total number of turns.

[0070] Changeover and debugging time Time taken from the issuance of the instruction to change workpiece specifications to the achievement of all evaluation indicators for three consecutive products.

[0071] Process parameter mapping accuracy: Using the offline optimal calibration parameters as the true value, calculate the accuracy given by GBDT (or manual experience). Pitch feedforward relative error And calculate the overall quality score Q.

[0072] Comparative Example Comparative Example 1: Complete Reproduction of Background Technology Existing open-loop, single-loop control: Uses passive tensioning guide wheel assembly to maintain tension, without servo torque closed loop; turntable operates at a constant angular velocity. Rotational operation lacks laser profile cross-section sensing, dual-channel non-circular compensation for rotational speed and feed; it lacks rotational visual inspection and defect feedback; axial feed is set in an open-loop manner with a fixed speed ratio, lacking an inner magnetic scale and an outer laser ranging ring; it lacks digital twin health monitoring and process self-learning, and changeover parameters are determined by operators through trial and error based on experience. This comparative example directly corresponds to the superimposed state of the five control defects described in the background technology: passive tension, constant speed winding, lack of online detection, open-loop spacing, and lack of self-optimization.

[0073] Comparative Example 2: Open-loop feed and passive tension only, with manual visual inspection: Based on Comparative Example 1, manual visual inspection after winding is added, but there is still no process correction and pitch closed loop, which is used to illustrate that post-inspection cannot replace process closed loop.

[0074] Comparative Example 3: Only by increasing the pitch open-loop fine adjustment under constant speed conditions, there is still no non-circular compensation and visual closed loop: The operator finely calibrated and fixed the feed rate ratio according to the empirical value of the circular cross section to illustrate that under the non-circular cross section, adjusting only the average pitch without adjusting the azimuth compensation still cannot eliminate the azimuth range of the overlay ratio.

[0075] Example Example 1: Enables half-square root compensation of laser contour dual channels, active tension closed loop, bidirectional correction of rotational vision, magnetic scale, dual-stage pitch closed loop of encoder inner ring and laser ranging outer ring, and digital twin health monitoring; the initial values ​​of process parameters are set to default values, and GBDT feedforward has not yet been enabled.

[0076] Example 2: Based on Example 1, a total of 8 batches of the same specification have been accumulated and GBDT retraining has been completed. Subsequent experiments will use the optimal process parameter mapping table output by process self-learning as the feedforward initial value for the tension control module and the feed control module.

[0077] Example 3: Based on Example 2, the number of quality image samples exceeds 500 and the defect trend prediction in Section 6.3 is enabled, allowing for intervention one cycle in advance.

[0078] Comparative experimental results Under the aforementioned uniform working conditions, the average results for 10 pieces in each group are as follows: In Comparative Example 1, the steady-state standard deviation of tension is approximately 0.52 N, the tension control accuracy is approximately 62%, the average absolute error of pitch is approximately 0.85 mm, the pitch control accuracy is approximately 41%, the azimuth range of the overlay ratio is approximately 28.0%, the defect outflow rate is approximately 6.8%, the changeover and debugging time is approximately 45 min, and the overall quality score is approximately 0.61.

[0079] In Comparative Example 2, after adding post-event manual visual inspection, the above indicators only improved slightly. The steady-state standard deviation of tension was about 0.51 N, the tension control accuracy was about 63%, the average absolute error of pitch was about 0.82 mm, the pitch control accuracy was about 43%, the azimuth range of the overlay ratio was about 27.5%, the defect outflow rate was about 5.9%, the changeover and debugging time was still about 45 minutes, and the overall quality score was about 0.63.

[0080] In Comparative Example 3, after only fine-tuning the open-loop setting of the average pitch, the average absolute error of the pitch decreased to about 0.46 mm, the pitch control accuracy increased to about 72%, but the azimuth range of the overlay ratio was still as high as about 22.0%, the tension control accuracy was about 64%, the defect outflow rate was about 4.1%, the changeover and debugging time was about 38 minutes, and the overall quality score was about 0.70.

[0081] After enabling full multi-loop coordination in Example 1, the steady-state standard deviation of tension decreased to about 0.08N, the tension control accuracy increased to about 98%, the average absolute error of pitch decreased to about 0.18mm, the pitch control accuracy increased to about 97%, the azimuth range of the overlay ratio decreased to about 4.0%, the defect outflow rate decreased to about 0.9%, the changeover and debugging time was shortened to about 26 minutes, and the overall quality score increased to about 0.91.

[0082] In Example 2, after further enabling GBDT process feedforward, the steady-state standard deviation of tension is about 0.07N, the tension control accuracy is about 99%, the average absolute error of pitch is about 0.15mm, the pitch control accuracy is about 98%, the azimuth range of the overlay ratio is about 3.6%, the defect outflow rate is about 0.7%, the changeover and debugging time is shortened to about 18 minutes, and the comprehensive quality score is about 0.94.

[0083] In Example 3, after enabling defect trend prediction, the defect outflow rate was further reduced to about 0.5%, and the other indicators were basically the same as in Example 2, with an overall quality score of about 0.95.

[0084] Regarding the accuracy of process parameter mapping, taking the offline calibration optimal value as the true value, in Comparative Example 1, when set manually based on experience, the relative error of tension feedforward is approximately 12.4%, the relative error of pitch / feed feedforward is approximately 15.8%, and the first convergence requires approximately 6.2 revolutions. In Example 1, under default parameters and online closed-loop conditions, the relative error of tension feedforward is approximately 6.8%, the relative error of pitch / feed feedforward is approximately 7.1%, and the first convergence requires approximately 2.8 revolutions. In Example 2, under GBDT feedforward and closed-loop conditions, the relative error of tension feedforward is approximately 5.7%, the relative error of pitch / feed feedforward is approximately 6.2%, and the first convergence requires approximately 1.6 revolutions. The relative errors are calculated as follows: , Similarly, calculations are performed based on pitch feedforward.

[0085] Comparative analysis results The comparative example shows large tension fluctuations and low accuracy. In Example 1, after enabling the servo torque closed loop, the steady-state standard deviation of tension decreased significantly and the accuracy improved, indicating that the active closed loop can effectively overcome the passive tensioning defects in the background technology.

[0086] In Comparative Example 1, the azimuth range of the lap ratio is very large. Even with fine adjustment of the average pitch in Comparative Example 3, the azimuth range is still too high, indicating that adjusting only the average amount without dual-channel azimuth compensation cannot solve the non-circular problem. In Example 1, after enabling half-square root rotational speeds and feed coupling, the azimuth range is reduced, demonstrating the advantage of dual-channel compensation in this application.

[0087] The comparative example showed a large pitch error and low accuracy; the example showed that after the two-stage closed loop, the average absolute pitch error was significantly reduced and the accuracy was improved, indicating that the inner ring of the magnetic scale and the outer ring of the laser ranging can effectively overcome the pitch defect of the open loop.

[0088] Although Comparative Example 2 increased post-event manual visual inspection, the improvement in defect outflow rate was limited. In Example 1, the defect outflow rate decreased significantly after bidirectional visual correction during the process. In Example 3, the defect outflow rate was further reduced after predictive intervention, proving that detection in the current loop and correction in the next loop have a significant advantage over post-event inspection.

[0089] Comparative Example 1 shows that manual experience results in large feedforward errors, slow initial convergence, and long changeover times. Example 2, after GBDT mapping, exhibits reduced relative feedforward errors, faster initial convergence, and shorter changeover times. Therefore, the improved data processing accuracy is specifically reflected in the fact that, under the same input feature definitions, the model outputs process parameters closer to the offline optimal true values, resulting in higher tension and pitch control accuracy and a higher overall quality score.

[0090] Comparative Example 3 only partially improves the open-loop calibration of the pitch, and many indicators are still inferior to the Example 1. The Example 1 improves simultaneously in tension, pitch, winding uniformity, defect outflow and changeover efficiency compared to Comparative Example 1, indicating that bidirectional data interaction and multi-loop coupling bring system-level gains, rather than the isolated effect of a single sensor.

[0091] In summary, the comparative examples reproduced the defects of the background technology, and the comparison results of the embodiments under the same working conditions can confirm the beneficial effects of this application, and specifically illustrate how the improvement of control accuracy and data processing accuracy can be reflected through verifiable indicators.

[0092] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An automatic wire harness branch tape winding device, comprising a main unit housing, a turntable assembly disposed inside the main unit housing, a first servo motor disposed inside the main unit housing and driving the turntable assembly to rotate via a gear transmission mechanism, and a main control unit disposed on the front of the main unit housing, wherein the turntable assembly includes a turntable and a platform extending to one side from the inner side of the turntable, a C-shaped wire passage opening is provided on one side of the turntable assembly and a tape positioning component is provided on one side of the turntable, the tape positioning component includes a guide cover disposed on one side of the turntable, a support plate is fixed on the guide cover, and a tape mounting shaft is rotatably disposed on one side of the support plate, characterized in that... Also includes: A wireless receiver module is installed on the fixed side of the main unit housing; The tension control module includes a servo tensioning roller assembly for performing closed-loop adjustment of belt tension and a servo torque sensor embedded in the support plate and cooperating therewith. The servo tensioning roller assembly includes: a tensioning roller, which is rotatably mounted on one side of the support plate; and a miniature servo motor, which is connected to the tensioning roller via a reducer. The tension control module is configured to calculate the extreme diameter of the cross section at each angle during one revolution of the turntable according to the extreme diameter formula of the elliptical cross section. The extreme diameter formula is the square root of the sum of the product of the major axis radius and the minor axis radius divided by the product of the square of the minor axis radius and the square of the cosine of the angle, plus the product of the square of the major axis radius and the square of the sine of the angle; the equivalent reference extreme diameter is taken as the square root of the product of the major axis radius and the minor axis radius. The tension control module generates a dynamic angular velocity compensation curve by multiplying the reference angular velocity by the square root of the ratio of the equivalent reference polar radius to the polar radius at that angle. The feed control module includes an axial feed mechanism for performing axial feed and pitch control of the wire harness; The profile triangulation measurement component includes multiple laser profile sensors evenly arranged on the fixed side of the turntable assembly to obtain cross-sectional geometric parameters, which are then input to the tension control module and the feed control module. The turntable dynamic angular velocity compensation data and differential feed speed compensation data are output respectively. Based on the turntable dynamic angular velocity compensation data and differential feed speed compensation data, the non-circular cross-section overlapping orientation deviation is eliminated through multi-loop coupling processing. The feed control module is configured to generate a differentiated feed rate curve by multiplying the base feed rate by the square root of the ratio of the extreme diameter at that angle to the equivalent reference extreme diameter. At the major axis position where the turntable rotation angle is zero, angular velocity compensation slows down the turntable, while feed rate compensation speeds up the feed. At the minor axis position where the turntable rotation angle is 90 degrees, angular velocity compensation speeds up the turntable, while feed rate compensation slows down the feed. The visual inspection module identifies defects in the image of the wrapped tape loop and obtains the identification results. The identification results are input to the tension control module and the feed control module, and the defect correction control command is output to correct the bidirectional deviation of tension and pitch in real time. The visual inspection module includes: a wireless image transmission module, which is embedded in one side of the table panel; a miniature industrial camera, which is embedded in the table panel by a fixing component; and a distance measuring sensor set on the other side of the table panel. An acoustic emission sensor is installed on the fixed side of the turntable assembly. Axial displacement data is obtained by a linear displacement sensor and combined with the absolute rotation angle data of the turntable. The data is then input to the feed control module for coupling processing to obtain the actual helical pitch data. The distance measuring joint sensor constructs a three-dimensional point cloud model of the contour of the wound tape based on the actual spiral pitch data, compares it with the preset standard contour to calculate the pitch deviation, and outputs a pitch deviation correction signal to feed the feed control module; the distance measuring joint sensor is set on the other side of the table panel and rotates synchronously with the turntable to align with the edge of the wound tape spiral. Acoustic emission signals and servo drive current signals of the spindle drive mechanism are collected by acoustic emission sensors and servo current sensors, respectively, and input to the spindle health monitoring link of the digital twin module.

2. The control system of the automatic winding device for wire harness branch tape according to claim 1, characterized in that: The digital twin module is connected to the tension control module, feed control module, contour triangulation measurement component and vision inspection module respectively, to acquire the full process operation parameters and defect detection data, and output the optimization parameters and control data in reverse, so as to realize the dynamic optimization of the winding process and the control of spindle faults. The guide cover has a through opening for the tape to pass through.

3. The control system of the automatic winding device for branched wire harness tape according to claim 2, characterized in that, The servo tensioning roller assembly includes a tensioning roller located between the belt mounting shaft and the guide cover, a rubber sleeve provided on the outside of the tensioning roller, a servo torque sensor located on the outside of the tensioning roller, a reducer fixed on one side of the support plate, and the reducer being axially connected to the tensioning roller. A micro servo motor is connected to the input end of the reducer. The micro servo motor controls the resistance torque applied to the conveyor belt by the tensioning roller by adjusting the output torque; The servo torque sensor detects the torque at the end of the tension roller drive shaft in real time, converts the shaft torque into a belt tension feedback value, and transmits it to the micro servo motor to form a tension closed loop. Multiple laser contour sensors perform contour scanning of the cross-section of the slender workpiece in a diameter direction, and the multiple contour scanning data are processed by point cloud fitting to obtain the cross-section type, major axis dimension, minor axis dimension and cross-section contour data of the slender workpiece. The tension control module adjusts the rotational speed of the turntable assembly angle by angle according to the cross-sectional geometric parameters, and the feed control module adjusts the axial feed speed of the slender workpiece.

4. The control system of the automatic winding device for wire harness branch tape according to claim 3, characterized in that: The turntable assembly includes a turntable rotatably disposed on its inner side, a cutter disposed on one edge of the table panel, and a C-shaped toothed disc disposed on the side of the turntable near the support plate. The visual inspection module also includes a replaceable battery pack disposed on one side of the support plate; The servo current sensor is installed on the output side of the first servo motor driver. The wireless image transmission module transmits the image data acquired by the miniature industrial camera and the real-time sampling data of the servo torque sensor to the wireless receiving module in a concurrent wireless data stream manner.

5. The control system of the automatic winding device for wire harness branch tape according to claim 4, characterized in that: It also includes an image processing and defect recognition unit integrated within the main control unit; The image processing and defect identification unit receives the image data stream forwarded by the wireless receiving module and performs online defect identification on each frame of the image. The identification objects include gap defects, stacking deviation defects, edge offset defects, and wrinkle defects. When an overlap deviation defect is detected, the actual pitch of the current cycle is calculated simultaneously, and the pitch deviation is transmitted to the feed control module as one of the defect correction signals to compensate for the feed speed of the next cycle; when a wrinkle or gap defect related to tension is detected, the tension deviation correction is transmitted to the tension control module as one of the defect correction signals to adjust the tension target value.

6. The control system of the automatic winding device for branched wire harness tape according to claim 5, characterized in that, The axial feed mechanism includes a motor disposed on one side thereon. A lead screw is rotatably disposed inside the axial feed mechanism and is axially connected to the output end of the motor. A wire harness clip is threaded to the outside of the lead screw. A slide is provided on the front of the axial feed mechanism and the wire harness clip is guided through the slide. The linear displacement sensor includes a magnetic scale, which is fixed along the inner side of the slide of the axial feed mechanism, and a wire harness clip connected to a reading head, which moves linearly along the magnetic scale with the wire harness clip. The main control unit reads the position value of the multi-turn absolute encoder built into the drive shaft of the first servo motor, and converts it according to the transmission ratio of the gear transmission mechanism to obtain the absolute rotation angle data of the turntable assembly. The ranging segment sensor independently verifies the actual helical pitch of the wound area online, and after comparing the verified pitch with the target pitch, performs outer loop compensation on the speed target value of the feed control module. The axial feed mechanism supports seamless online switching between three pitch modes: interval winding, half-overlap winding, and full-overlap winding.

7. The control system of the automatic winding device for wire harness branch tape according to claim 6, characterized in that: The digital twin module includes a digital twin engine and a self-learning optimization module, and receives all process data from the tension control module, the feed control module and the vision detection module; The digital twin engine performs wavelet packet time-frequency decomposition on the ultrasonic signal collected by the acoustic emission sensor, extracts the sub-band energy envelope feature value and compares it with the baseline mean plus three times the standard deviation threshold established by the initial health data of the equipment. When the feature value exceeds the threshold, a maintenance warning is triggered and a health control command to reduce the turntable speed is output. The servo current sensor collects the three-phase current of the first servo motor in real time and converts it into an estimated load torque value. When the load torque continuously exceeds 1.2 times the average value of the baseline under normal operating conditions and the duration exceeds 3 seconds, an abnormal load warning is triggered. The self-learning optimization module uses workpiece specifications as the primary key and inputs the input features, consisting of cross-sectional geometry sub-vectors, tension process sub-vectors, feed process sub-vectors, and visual quality sub-vectors, into a gradient boosting decision tree regression model. The regression model is ensembled from multiple CART regression trees of depth 4 to 6 using an additive learning rate of 0.05 to 0.1, and trained using multi-output weighted mean square error as the loss function. Training is triggered when there are at least 5 historical batches of the same specification. The output includes a process parameter mapping table containing the optimal tension target value, optimal reference angular velocity, and optimal reference feed speed, which is then sent back to the tension control module and the feed control module as the initial control parameters for subsequent batches of the same specification.

8. The control method of the control system for the automatic winding device for branched wire harness tape according to claim 7, characterized in that, Includes the following steps: S1. The cross-sectional geometric parameters are obtained by the contour triangulation measurement component and simultaneously output to the tension control module and the feed control module; S2. Generate dynamic angular velocity compensation curves and differentiated feed rate curves according to the extreme diameter formula and the square root distribution logic of each half compensation, so that the two compensations can eliminate the non-circular cross section overlay ratio azimuth deviation after multi-loop coupling. S3. The tension control module actively adjusts the tape tension in a closed loop based on the feedback from the servo torque sensor, and drives the turntable assembly to rotate and wind according to the dynamic angular velocity compensation curve. S4. The visual inspection module performs image acquisition and defect identification, and simultaneously outputs defect correction signals to the tension control module and the feed control module. S5. The feed control module couples the axial displacement data of the magnetic scale with the absolute rotation angle data of the turntable to form a closed loop of the pitch inner loop, and receives the outer loop verification value of the distance measuring segment sensor for compensation. S6. The digital twin module aggregates process data, outputs health control instructions and optimal process parameter mapping table, and sends them back in reverse.

9. The control method according to claim 8, characterized in that: In step S3, the closed-loop control accuracy of the tape tension is not less than ±0.1N, and the sampling period of the main controller is not greater than 1ms.

10. The control method according to claim 8, characterized in that: In step S4, four types of defects—gap, overlapping deviation, edge offset, and wrinkles—are identified in parallel for each winding image. When an overlapping deviation defect is identified, the actual pitch of this winding is calculated simultaneously, and the pitch deviation is added to the inner ring error integral term of the magnetic scale of the feed control module to compensate for the feed speed of the next winding. In step S6, the frequency band energy envelope of the acoustic emission sensor signal is extracted by wavelet packet time-frequency decomposition and compared with the baseline mean plus three times the standard deviation threshold. The health judgment is made in combination with the load torque calculated by the servo current sensor. The input features consisting of cross-sectional geometry, tension process, feed process and visual quality are input into the trained gradient boosting decision tree regression model, and the optimal process parameter mapping table is output.

Images