Motor stator iron sheet angle positioning method and system

Abstract

The invention relates to the technical field of motor stator iron sheet positioning, in particular to a motor stator iron sheet angle positioning method and system, and the method comprises the following steps: S1, collecting real-time images of T-shaped stator iron sheets which are disorderly arranged on a conveying belt through a visual sensor; s2, identifying the features of the T-shaped end part and the center hole of the iron sheet based on a deep learning model; the feature recognition comprises the following steps: positioning a minimum enclosing rectangular frame of an iron sheet by adopting a YOLOv5 model; extracting a T-shaped end contour and a central hole edge through a U-Net segmentation network; calculating a contour centroid as a feature reference point; s3, calculating the actual angle offset alpha of the iron sheet according to the coordinates of the feature points; and S4, the multi-degree-of-freedom mechanical arm is controlled to grab the iron sheet at the compensation angle alpha and transfer the iron sheet to an assembly station. The assembling problem caused by angle deviation of the iron sheet can be effectively reduced, the assembling quality and stability of the motor stator iron core are improved, the defective rate is reduced, the assembling speed is increased, and the production efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of motor stator sheet positioning technology, and in particular to a method and system for angular positioning of motor stator sheets. Background Technology

[0002] As a widely used power device in industrial production and daily life, the performance of electric motors directly affects the operation of various electromechanical products. The motor stator, as a key component of the motor, is assembled from a large number of stator iron sheets, and the assembly precision of these stator iron sheets plays a decisive role in the performance and stability of the motor.

[0003] In the traditional assembly process of motor stator laminations, the positioning step is crucial for ensuring assembly accuracy. Existing stator lamination positioning technologies mainly rely on mechanical clamps and simple vision recognition systems. While mechanical clamps can fix the lamination position to a certain extent, factors such as manufacturing errors, wear of mechanical components, and loosening during installation make it difficult to guarantee high precision and consistency in positioning every time. Simple vision recognition systems are limited by the resolution of image acquisition equipment, the accuracy of processing algorithms, and interference from environmental factors, resulting in significant errors in the identification and positioning of lamination features.

[0004] Insufficient positioning accuracy directly leads to stacking deviations in the stator sheets during subsequent assembly, making it difficult to meet design requirements for key indicators such as the concentricity and perpendicularity of the motor stator. Therefore, new improvements are needed to address the existing stator sheet positioning. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method and system for precisely positioning motor stator iron sheets. This method effectively reduces assembly problems caused by angle deviations in the iron sheets, improves the assembly quality and stability of motor stator cores, reduces the defect rate, accelerates assembly speed, and increases production efficiency.

[0006] The technical solution adopted in this invention is: a method for angular positioning of motor stator iron sheets, characterized by the following steps: Step S1: Acquire real-time images of the disordered T-shaped stator iron sheets on the conveyor belt using a vision sensor; Step S2: Identify the T-shaped end and center hole features of the iron sheet based on a deep learning model; feature identification includes: using the YOLOv5 model to locate the minimum bounding rectangle of the iron sheet; extracting the T-shaped end contour and the edge of the center hole through the U-Net segmentation network; and calculating the centroid of the contour as the feature reference point. Step S3: Calculate the actual angular offset α of the iron sheet based on the coordinates of the feature points. The angle calculation includes: establishing a polar coordinate system with the center of the central hole as the origin; calculating the theoretical angle β=arctan based on the coordinates of the centroid of the T-shaped end of the stator iron sheet; the actual offset α=β-θ, where θ is the angle between the direction of the conveyor belt movement and the X-axis of the coordinate system. Step S4: Control the multi-degree-of-freedom robotic arm to grab the iron sheet at a compensation angle α and transfer it to the assembly station; the gripping control includes: installing an electromagnetic chuck at the end of the robotic arm; using the RRT algorithm for path planning to avoid obstacles on the conveyor belt; applying a rotation compensation amount α during gripping to bring the angle of the iron sheet to zero.

[0007] A further improvement to the above scheme is that, in step S2, the deep learning model is trained using: Data augmentation: Generate a composite image of the iron sheet that includes changes in lighting and occlusion; Loss function: ,in ; Transfer learning: Fine-tuning of weights pre-trained on the COCO dataset.

[0008] A further improvement to the above scheme is the inclusion of a correction step: A laser displacement sensor is installed at the assembly station; The actual angular deviation δ of the iron sheet is detected. If δ > 0.5°, a secondary calibration is triggered. After calibration, the deviation is written into the historical database to optimize the angle prediction model.

[0009] A further improvement to the above scheme is that, in step S1, vibration suppression of the conveyor belt is required. Vibration suppression of the conveyor belt includes: triggering the conveyor belt brake during the exposure period of the visual sensor; braking duration Δt = 1 / 2f, where f is the dominant frequency of the conveyor belt vibration; obtaining the dominant frequency f includes: collecting the conveyor belt vibration signal through an accelerometer; and determining the peak value of the dominant frequency based on FFT spectrum analysis.

[0010] A further improvement to the above scheme is that the stacking accuracy for the stator rings is as follows: single-layer angular error ≤ 0.3°; cumulative error of 100 layers ≤ 1.5°.

[0011] A stator sheet angle positioning system is disclosed for implementing a method for positioning motor stator sheets at angles. The system includes a vibration-suppressing conveying device, an industrial camera, a processing unit, and a six-axis robotic arm. The vibration-suppressing conveying device is used for conveying T-shaped stator sheets and is equipped with a brake and an acceleration sensor. The industrial camera includes a ring light source and a polarizing filter. The processing unit deploys a feature recognition model and an angle calculation algorithm. The six-axis robotic arm has an electromagnetic chuck and a force sensor at its end.

[0012] A further improvement to the above solution is that the industrial camera meets the following requirements: resolution ≥ 5MP, frame rate 120fps; installation height H = 500±10mm, tilt angle γ = 15°; ring light source color temperature 5600K, brightness adjustable.

[0013] A further improvement to the above scheme is that the processing unit includes: The feature recognition module runs YOLOv5 and U-Net models; the input image size is 640×640 pixels; the inference latency is ≤8ms; and the feature localization accuracy is ±0.1 pixels. The motion control module generates the trajectory and angle compensation commands for the six-axis robotic arm. The data storage module records historical deviation data.

[0014] A further improvement to the above solution is that the end effector of the six-axis robotic arm includes: an electromagnetic chuck with adjustable suction force of 0~50N; a six-dimensional force sensor with a range of ±200N; and a miniature rotary motor with a resolution of 0.01°.

[0015] A further improvement to the above scheme is the inclusion of a secondary calibration unit: Laser displacement sensor to measure the height difference at the edges of a metal sheet; Calibration algorithm: If the height difference Δh > 10 μm, then calculate the compensation angle. , where D is the diameter of the iron sheet.

[0016] A further improvement to the above solution is that the industrial camera and the processing unit communicate via the GigE Vision protocol; The processing unit is synchronized with the six-axis robotic arm via an EtherCAT bus; Real-time data transmission latency ≤1ms.

[0017] The beneficial effects of the present invention are: Compared to existing stator sheet assembly and positioning methods, this invention, in the image acquisition stage, uses a vision sensor to acquire real-time images of the randomly arranged T-shaped stator sheets on the conveyor belt, enabling timely acquisition of sheet status information. This provides a foundation for subsequent precise positioning, ensuring that operations are performed based on the latest sheet distribution throughout the positioning process, improving the timeliness and accuracy of positioning. The feature recognition step employs an advanced deep learning model. The YOLOv5 model is used to locate the minimum bounding rectangle of the sheet, quickly and accurately pinpointing its position and defining the range for subsequent feature extraction. The U-Net segmentation network extracts the T-shaped end contour and the edge of the central hole, further refining the feature information. The centroid of the contour is calculated as the feature reference point, clarifying the key reference point for positioning. This series of operations, combined, can accurately identify the features of the T-shaped end and the central hole of the sheet, providing reliable data support for angle calculation. In the angle calculation stage, a polar coordinate system is established with the center of the central hole as the origin. The theoretical angle is calculated based on the coordinates of the centroid of the T-shaped end, and the actual offset is calculated by combining the angle between the conveyor belt's movement direction and the X-axis of the coordinate system. This coordinate-based calculation method can accurately measure the actual angular deviation of the iron sheet, making the positioning more scientific and accurate. In terms of gripping control, an electromagnetic chuck is installed at the end of the robotic arm to reliably grip the iron sheet; the path planning uses the RRT algorithm to avoid obstacles on the conveyor belt, ensuring the robotic arm operates safely and efficiently; during gripping, a rotational compensation is applied to bring the iron sheet angle to zero, ensuring that the iron sheet is transferred to the assembly station at the correct angle. This embodiment, through precise positioning, can effectively reduce assembly problems caused by iron sheet angular deviations, improve the assembly quality and stability of the motor stator core, reduce the defect rate, accelerate assembly speed, and increase production efficiency.

[0018] A stator sheet angle positioning system is provided for implementing a method of positioning motor stator sheets at angles. The vibration-suppressing conveyor effectively ensures stable stator sheet conveying. The combination of a brake and an acceleration sensor allows for real-time monitoring of vibration during conveying and timely braking and adjustment to prevent changes in sheet position due to vibration. This provides a stable sheet for subsequent image acquisition, improving image quality and the accuracy of subsequent feature recognition. An industrial camera equipped with a ring light source and a polarizing filter optimizes image acquisition. The ring light source provides uniform illumination and reduces shadow interference; the polarizing filter reduces reflections, resulting in clearer real-time images of the acquired T-shaped stator sheets, which is beneficial for subsequent feature recognition based on deep learning models. The processing unit deploys a feature recognition model and angle calculation algorithm. Utilizing advanced deep learning models and precise angle calculation methods, it can quickly and accurately identify sheet features and calculate angular offsets, ensuring high positioning accuracy and efficiency. An electromagnetic chuck at the end of a six-axis robotic arm reliably grips the sheet, and a force sensor monitors the gripping force in real time, preventing damage to the sheet during gripping. The path planning employs the RRT algorithm to avoid obstacles on the conveyor belt. Simultaneously, rotational compensation is applied during gripping to bring the angle of the iron sheet to zero, ensuring accurate transfer of the iron sheet to the assembly station. This invention provides reliable technical support for the angle positioning of stator iron sheets before motor stator core assembly, effectively improving positioning accuracy and efficiency, thereby enhancing the assembly quality of the motor stator core. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the method for angular positioning of motor stator plates according to the present invention. Figure 2 This is a schematic diagram of the stator sheet angle positioning system of the present invention; Figure 3 for Figure 2 A schematic diagram of the stator sheet angle positioning system from another perspective; Figure 4 for Figure 2 A schematic diagram of the connection of the stator sheet angle positioning system.

[0020] Explanation of reference numerals in the attached drawings: 1. Vibration suppression conveyor device; 11. Brake; 12. Accelerometer; 2. Industrial camera; 21. Ring light source; 22. Polarizing filter; 3. Six-axis robotic arm; 31. Electromagnetic chuck; 32. Six-dimensional force sensor; 33. Miniature rotary motor; 4. Processing unit; 41. Feature recognition module; 42. Motion control module; 43. Data storage module; 5. Secondary calibration unit; 51. Laser displacement sensor. Detailed Implementation

[0021] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0022] It should be noted that when a component is said to be "fixed to" another component, it can be directly attached to the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. Figure 1 As shown, in one embodiment of the present invention, a method for angular positioning of motor stator plates is provided, comprising the following steps: Step S1: Acquire real-time images of the disordered T-shaped stator iron sheets on the conveyor belt using a vision sensor; Step S2: Identify the T-shaped end and center hole features of the iron sheet based on a deep learning model; feature identification includes: using the YOLOv5 model to locate the minimum bounding rectangle of the iron sheet; extracting the T-shaped end contour and the edge of the center hole through the U-Net segmentation network; and calculating the centroid of the contour as the feature reference point. Step S3: Calculate the actual angular offset α of the iron sheet based on the coordinates of the feature points. The angle calculation includes: establishing a polar coordinate system with the center of the central hole as the origin; calculating the theoretical angle β=arctan based on the coordinates of the centroid of the T-shaped end of the stator iron sheet; the actual offset α=β-θ, where θ is the angle between the direction of the conveyor belt movement and the X-axis of the coordinate system. Step S4: Control the multi-degree-of-freedom robotic arm to grab the iron sheet at a compensation angle α and transfer it to the assembly station; the gripping control includes: installing an electromagnetic chuck at the end of the robotic arm; using the RRT algorithm for path planning to avoid obstacles on the conveyor belt; applying a rotation compensation amount α during gripping to bring the angle of the iron sheet to zero.

[0024] In this embodiment, during the image acquisition stage, a vision sensor captures real-time images of the randomly arranged T-shaped stator iron sheets on the conveyor belt, enabling timely acquisition of the sheet status information. This provides a foundation for subsequent precise positioning, ensuring that operations are performed based on the latest iron sheet distribution throughout the positioning process, improving the timeliness and accuracy of positioning. The feature recognition step employs an advanced deep learning model. The YOLOv5 model is used to locate the minimum bounding rectangle of the iron sheet, quickly and accurately pinpointing its position and defining the range for subsequent feature extraction. The U-Net segmentation network extracts the T-shaped end contour and the edge of the central hole, further refining the feature information. The centroid of the contour is calculated as the feature reference point, clarifying the key reference point for positioning. This series of operations, combined, accurately identifies the features of the T-shaped end and the central hole of the iron sheet, providing reliable data support for angle calculation. In the angle calculation stage, a polar coordinate system is established with the center of the central hole as the origin. The theoretical angle is calculated based on the coordinates of the centroid of the T-shaped end, and the actual offset is calculated by combining the angle between the conveyor belt's movement direction and the X-axis of the coordinate system. This coordinate-based calculation method accurately measures the actual angular offset of the iron sheet, making positioning more scientific and accurate. In terms of gripping control, an electromagnetic chuck is installed at the end of the robotic arm, enabling reliable gripping of the iron sheet. Path planning employs the RRT algorithm to avoid obstacles on the conveyor belt, ensuring safe and efficient operation of the robotic arm. During gripping, rotational compensation is applied to bring the iron sheet angle to zero, ensuring the iron sheet is transferred to the assembly station at the correct angle. This embodiment, through precise positioning, effectively reduces assembly problems caused by iron sheet angle deviations, improves the assembly quality and stability of the motor stator core, reduces the defect rate, accelerates assembly speed, and increases production efficiency.

[0025] The above embodiments are implemented in the environment shown in the table below:

[0026] The algorithm tests of the above embodiments and the comparison of feature recognition accuracy are shown in the table below:

[0027] In the table above, the YOLOv5 model has an input size of 640×640 and a confidence threshold of 0.95; the U-Net segmentation uses a ResNet34 backbone network; and the centroid calculation error after feature fusion is ≤ ±0.1 pixels.

[0028] The following table compares the angle positioning of the above embodiments under different working conditions:

[0029] The cumulative error of the stacking is shown in the table below:

[0030] The motion parameters of the gripping control robotic arm in the above embodiment are shown in the table below:

[0031] The above embodiments were tested in continuous 8-hour production, as shown in the table below:

[0032] In step S2, the training of the deep learning model employs: data augmentation: generating a synthetic image of the iron sheet including illumination changes and occlusions; loss function: ,in Transfer learning: Pre-trained weight fine-tuning based on the COCO dataset. In this embodiment, for deep learning model training, data augmentation generates synthetic images of iron sheets containing lighting changes and occlusions, greatly enriching the diversity of training data. This enables the model to adapt to various complex real-world scenarios, accurately identifying the T-shaped ends and central hole features of the iron sheets even in poor lighting or with partial occlusion, effectively improving the robustness and accuracy of feature recognition. Using a specific loss function helps the model continuously optimize parameters during training, making the model's predictions closer to the true values, thereby improving the overall performance of the model. Transfer learning, based on pre-trained weight fine-tuning on the COCO dataset, fully utilizes the knowledge of the existing dataset, reducing model training time and computational resource consumption, while accelerating model convergence, enabling the model to achieve better recognition results more quickly.

[0033] The system also includes a correction step: a laser displacement sensor is installed at the assembly station; the actual angular deviation δ of the iron sheet is detected; if δ > 0.5°, a secondary calibration is triggered; the calibrated deviation is written into the historical database to optimize the angle prediction model. In this embodiment, the addition of the correction step further improves the positioning accuracy. The laser displacement sensor at the assembly station can detect the actual angular deviation of the iron sheet in real time. When the deviation exceeds 0.5°, a secondary calibration is triggered to ensure that the angle of the iron sheet meets the requirements before assembly. Writing the calibrated deviation into the historical database can be used to optimize the angle prediction model, enabling the model to more accurately predict the angle of the iron sheet in subsequent positioning processes, forming a continuously optimizing closed-loop system.

[0034] The following table shows the angle changes before and after the second calibration in this embodiment (tested on 500 pieces):

[0035] Under the above conditions, the calibration agency's maximum compensation capability is ±2.0°; after calibration, all iron pieces meet the assembly standard of δ≤0.15°.

[0036] In step S1, vibration suppression of the conveyor belt is required. This includes: triggering the conveyor belt brake during the visual sensor exposure period; braking duration Δt = 1 / 2f, where f is the dominant vibration frequency of the conveyor belt; obtaining the dominant vibration frequency f includes: acquiring the conveyor belt vibration signal through an accelerometer; and determining the peak value of the dominant frequency based on FFT spectrum analysis. In this embodiment, triggering the conveyor belt brake during the visual sensor exposure period effectively reduces the vibration displacement of the iron sheet during exposure. A stable state of the iron sheet is crucial during image acquisition. Vibration can blur the acquired image and cause inaccurate positional information, thus affecting subsequent feature recognition and angle calculation. By braking during the exposure period, it is ensured that the visual sensor acquires clear and accurate real-time images of the T-shaped stator iron sheet, providing a high-quality data foundation for precise positioning. The braking duration Δt is set to 1 / 2f, where f is the dominant vibration frequency of the conveyor belt, which is the optimal braking time derived from scientific calculations. Acquiring the conveyor belt vibration signal through an accelerometer and determining the peak value of the dominant frequency based on FFT spectrum analysis allows for precise understanding of the conveyor belt's vibration characteristics. A reasonable braking duration effectively suppresses vibration without excessively affecting the normal conveyor belt efficiency, achieving a balance between vibration suppression and production efficiency. Accurate vibration suppression helps improve the accuracy of feature recognition. Stable images enable deep learning models to more precisely locate the minimum bounding rectangle of the iron sheet, extract the T-shaped end contour and the edge of the center hole, and calculate more accurate feature reference points. This further improves the accuracy of angle calculation, making the calculated actual offset α closer to the true value.

[0037] For the stacking assembly of stator iron rings, the stacking accuracy must meet the following requirements: single-layer angle error ≤ 0.3°; cumulative error per hundred layers ≤ 1.5°. In this embodiment, regarding the single-layer angle error, a single-layer angle error of ≤ 0.3° is required, making the angle positioning method more accurate in the angle positioning of each stator iron piece. Whether it's image acquisition by visual sensors, feature recognition by deep learning models, calculation of angle offsets, or control of the robotic arm's grasping, all stages must be strictly controlled to ensure that the angle deviation of each iron piece is within a minimal range. This effectively reduces the adverse effects on motor performance caused by inaccurate angles of single iron pieces. For the cumulative error per hundred layers, a cumulative error of ≤ 1.5° is required, meaning that the angle positioning method must possess high stability and consistency during the stacking of a large number of stator iron pieces. Continuous optimization and calibration are necessary to avoid error accumulation. High-precision angle positioning ensures the quality of the stator iron ring stacking, making the stator core performance more stable and reliable. This is implemented using a motor stator iron piece angle positioning method.

[0038] like Figures 1-4As shown, a stator sheet angle positioning system is used to implement a method for positioning the angle of motor stator sheets. It includes a vibration-suppressing conveying device 1, an industrial camera 2, a processing unit 4, and a six-axis robotic arm 3. The vibration-suppressing conveying device 1 is used for conveying T-shaped stator sheets and is equipped with a brake 11 and an acceleration sensor 12. The industrial camera 2 includes a ring light source 21 and a polarizing filter 22. The processing unit 4 deploys a feature recognition model and an angle calculation algorithm. The six-axis robotic arm 3 has an electromagnetic chuck 31 and a force sensor 32 at its end. This embodiment is used to implement a method for positioning the angle of motor stator sheets. The vibration-suppressing conveying device 1 effectively ensures the stable conveying of the stator sheets. The combination of the brake 11 and the acceleration sensor 12 can monitor the vibration during the conveying process in real time and adjust the brake in time to avoid changes in the sheet position due to vibration, providing a stable sheet for subsequent image acquisition, improving image quality and the accuracy of subsequent feature recognition. The industrial camera 2, equipped with a ring light source 21 and a polarizing filter 22, optimizes the image acquisition effect. The ring light source 21 provides uniform illumination and reduces shadow interference; the polarizing filter 22 reduces the impact of reflection, making the real-time image of the acquired T-shaped stator iron sheet clearer, which is beneficial for subsequent feature recognition based on deep learning models. The processing unit 4 deploys feature recognition models and angle calculation algorithms. With the help of advanced deep learning models and precise angle calculation methods, it can quickly and accurately identify the iron sheet features and calculate the angle offset, ensuring high positioning accuracy and efficiency. The electromagnetic chuck 31 at the end of the six-axis robotic arm 3 can reliably grasp the iron sheet, and the force sensor 32 can monitor the grasping force in real time to avoid damage to the iron sheet during the grasping process. The path planning adopts the RRT algorithm to avoid obstacles on the conveyor belt. At the same time, a rotation compensation is applied during grasping to bring the iron sheet angle to zero, ensuring accurate transfer of the iron sheet to the assembly station. This invention provides reliable technical support for the angle positioning of stator iron sheets before the assembly of motor stator cores, effectively improving positioning accuracy and efficiency, and thus improving the assembly quality of motor stator cores.

[0039] The industrial camera 2 meets the following requirements: resolution ≥ 5MP, frame rate 120fps; installation height H = 500±10mm, tilt angle γ = 15°; ring light source 21 with a color temperature of 5600K and adjustable brightness. In this embodiment, the resolution ≥ 5MP and frame rate 120fps allow for the capture of finer features of the T-shaped stator sheet, providing rich and accurate image information for subsequent feature recognition and angle calculation. This enables the processing unit 4 to more accurately identify the feature points of the sheet and improve the accuracy of angle positioning. The high frame rate allows for the acquisition of multiple frames in a short time, meeting the needs of rapid positioning and improving the system's efficiency. The precise settings of the installation height H = 500±10mm and tilt angle γ = 15° ensure that the industrial camera 2 can capture images of the stator sheet from a suitable angle and distance, reducing image distortion and projection errors. This ensures that the acquired images accurately reflect the actual state of the sheet, which is beneficial for accurate analysis by the feature recognition model. The ring light source features a 21-color-temperature (5600K) ring with adjustable brightness. The 5600K color temperature closely approximates natural light, providing a uniform and natural lighting environment and reducing the impact of shadows and reflections on image quality. Adjustable brightness allows for flexible adjustments based on different working scenarios and the characteristics of the metal sheet, further optimizing image acquisition and ensuring image clarity and contrast.

[0040] Processing unit 4 includes: a feature recognition module 41, which runs YOLOv5 and U-Net models; input image size 640×640 pixels; inference latency ≤8ms; feature localization accuracy ±0.1 pixels; a motion control module 42, which generates a six-axis robotic arm trajectory and angle compensation commands; and a data storage module 43, which records historical deviation data. In this embodiment, the feature recognition module 41 runs YOLOv5 and U-Net models, with an input image size of 640×640 pixels, an inference latency ≤8ms, and a feature localization accuracy of ±0.1 pixels. The powerful recognition capabilities of YOLOv5 and U-Net, combined with a suitable image size, can quickly and accurately identify the T-shaped end and center hole features of the iron sheet. The extremely short inference latency ensures that the system can respond promptly and process image data efficiently. High-precision feature localization provides a solid foundation for accurately calculating the angle of the iron sheet, reducing angle localization deviations caused by feature recognition errors. The motion control module 42 can generate a six-axis robotic arm trajectory and angle compensation commands. Based on the angle information obtained from the feature recognition module 41, the robotic arm's movements can be precisely planned, enabling it to accurately grasp and place the iron sheet. Simultaneously, real-time compensation is performed based on angle deviations to ensure the iron sheet is accurately installed in the designated position. The data storage module 43 records historical deviation data, which can be used to analyze and optimize the angle positioning system. By mining and learning from historical deviation data, the feature recognition model and motion control strategy can be further adjusted, continuously improving the system's positioning accuracy and stability, making the motor stator iron sheet angle positioning system more intelligent and reliable.

[0041] The six-axis robotic arm's end effector includes: an electromagnetic chuck 31 with adjustable suction force from 0 to 50 N; a six-dimensional force sensor 32 with a range of ±200 N; and a miniature rotary motor 33 with a resolution of 0.01°. In this embodiment, the electromagnetic chuck 31's adjustable suction force from 0 to 50 N allows the robotic arm to flexibly adjust the suction force according to the different materials, sizes, and weights of the stator iron sheets. For lighter, more fragile iron sheets, a smaller suction force can be used to avoid damage caused by excessive suction; for heavier iron sheets, sufficient suction force is provided to ensure stable gripping and handling, greatly improving the system's applicability to different iron sheets and operational safety. The six-dimensional force sensor 32, with a range of ±200 N, can monitor the force applied to the robotic arm during the gripping and placement of the iron sheets in real time. When the iron sheet contacts the target position, the sensor can accurately sense the magnitude and direction of the force. This helps to promptly detect potential collisions or uneven force distribution, thereby enabling precise force feedback control and ensuring the stability and accuracy of the iron sheet during installation. The miniature rotary motor 33 has a resolution of 0.01°, enabling high-precision angle fine-tuning. After the processing unit 4 calculates the actual offset of the iron sheet, the miniature rotary motor 33 can precisely rotate the iron sheet with extremely small angle increments according to the angle compensation command, effectively correcting the angle deviation and improving the accuracy of the angle positioning of the iron sheet on the motor stator.

[0042] It also includes a secondary calibration unit 5, which includes a laser displacement sensor 51 for measuring the height difference at the edge of the iron sheet; Calibration algorithm: If the height difference Δh > 10 μm, then calculate the compensation angle. Where D is the diameter of the iron sheet. In this embodiment, the laser displacement sensor 51 can measure the height difference at the edge of the iron sheet, accurately capturing subtle height changes that may occur during the stacking process. In actual operation, even if the initial angle positioning has reached a certain accuracy, various factors, such as uneven iron sheet surface and minor vibrations during installation, may still cause height differences at the edge of the iron sheet. The laser displacement sensor 51, with its high-precision measurement capability, can detect these subtle changes in a timely manner. The application of the calibration algorithm enables the system to accurately compensate for the height difference. When the height difference Δh > 10μm, the compensation angle is calculated using a specific formula. This ensures that when a large height difference occurs, the system will not ignore these deviations but can actively adjust them. Based on the calculated compensation angle, the motion control module 42 can control the six-axis robotic arm 3 to fine-tune the angle of the iron sheet, further correcting the angle deviation. The secondary calibration unit 5 improves the accuracy and reliability of the angle positioning of the motor stator iron sheet, reduces the angle error caused by the height difference, and ensures the high quality of the stator iron ring stacking assembly.

[0043] The industrial camera 2 communicates with the processing unit 4 via the GigE Vision protocol; the processing unit 4 synchronizes with the six-axis robotic arm via the EtherCAT bus. In this embodiment, the GigE Vision protocol has high-speed data transmission capabilities, enabling the rapid and stable transmission of high-resolution, high-frame-rate images acquired by the industrial camera 2 to the processing unit 4. The large amount of image data generated by the industrial camera 2, with a resolution ≥ 5MP and a frame rate of 120fps, can be transmitted without delay using this protocol, ensuring that the processing unit 4 can acquire the latest image information in a timely manner. This provides a data foundation for the efficient operation of the feature recognition module 41, thereby quickly and accurately identifying the features of the stator iron sheet and achieving angle calculation and positioning. EtherCAT bus synchronization ensures efficient collaboration between the processing unit 4 and the robotic arm. The angle positioning results and compensation commands derived from image analysis by the processing unit 4 can be quickly transmitted to the robotic arm via the EtherCAT bus. The bus has extremely low communication latency and high synchronization, enabling the robotic arm to respond to commands in real time and accurately execute the actions of grasping, rotating, and placing the stator iron sheet.

[0044] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for angular positioning of motor stator plates, characterized in that: Includes the following steps: Step S1: Acquire real-time images of the disordered T-shaped stator iron sheets on the conveyor belt using a vision sensor; Step S2: Identify the T-shaped ends and central hole features of the iron sheet based on a deep learning model; Feature recognition includes: locating the minimum bounding rectangle of the iron sheet using the YOLOv5 model; extracting the T-shaped end contour and the edge of the central hole using the U-Net segmentation network; and calculating the centroid of the contour as the feature reference point. Step S3: Calculate the actual angular offset α of the iron sheet based on the coordinates of the feature points. The angle calculation includes: establishing a polar coordinate system with the center of the central hole as the origin; calculating the theoretical angle β=arctan based on the coordinates of the centroid of the T-shaped end of the stator iron sheet; the actual offset α=β-θ, where θ is the angle between the direction of the conveyor belt movement and the X-axis of the coordinate system. Step S4: Control the multi-degree-of-freedom robotic arm to grab the iron sheet at a compensation angle α and transfer it to the assembly station; the gripping control includes: installing an electromagnetic chuck at the end of the robotic arm; using the RRT algorithm for path planning to avoid obstacles on the conveyor belt; applying a rotation compensation amount α during gripping to bring the angle of the iron sheet to zero.

2. The method for angular positioning of motor stator plates according to claim 1, characterized in that: In step S2, the deep learning model is trained using: Data augmentation: Generate a composite image of the iron sheet that includes changes in lighting and occlusion; Loss function: ,in ; Transfer learning: Fine-tuning of weights pre-trained on the COCO dataset.

3. The method for angular positioning of motor stator plates according to claim 1, characterized in that: It also includes corrective steps: A laser displacement sensor is installed at the assembly station; The actual angular deviation δ of the iron sheet is detected. If δ > 0.5°, a secondary calibration is triggered. After calibration, the deviation is written into the historical database to optimize the angle prediction model.

4. The method for angular positioning of motor stator plates according to claim 1, characterized in that: In step S1, vibration suppression of the conveyor belt is required. Vibration suppression of the conveyor belt includes: triggering the conveyor belt brake during the exposure period of the visual sensor; braking duration Δt = 1 / 2f, where f is the dominant vibration frequency of the conveyor belt; obtaining the dominant vibration frequency f includes: collecting the vibration signal of the conveyor belt through an accelerometer; and determining the peak value of the dominant frequency based on FFT spectrum analysis.

5. The method for angular positioning of motor stator plates according to claim 1, characterized in that: For the stacking assembly of stator iron rings, the stacking accuracy must meet the following requirements: single-layer angular error ≤ 0.3°; cumulative error of 100 layers ≤ 1.5°.

6. A stator sheet angle positioning system, used to implement the motor stator sheet angle positioning method according to any one of claims 1 to 5, characterized in that, include: A vibration-suppressing conveyor device is used for conveying T-shaped stator iron sheets, and the vibration-suppressing conveyor device is equipped with a brake and an acceleration sensor. An industrial camera, the industrial camera including a ring light source and a polarizing filter; A processing unit, wherein the processing unit deploys a feature recognition model and an angle calculation algorithm; and A six-axis robotic arm, wherein an electromagnetic chuck and a force sensor are provided at the end of the six-axis robotic arm.

7. The motor stator sheet angle positioning system according to claim 6, characterized in that: The industrial camera meets the following requirements: resolution ≥ 5MP, frame rate 120fps; installation height H = 500±10mm, tilt angle γ = 15°; ring light source color temperature 5600K, brightness adjustable.

8. The motor stator sheet angle positioning system according to claim 6, characterized in that: The processing unit includes: The feature recognition module runs YOLOv5 and U-Net models; the input image size is 640×640 pixels; the inference latency is ≤8ms; and the feature localization accuracy is ±0.1 pixels. The motion control module generates the trajectory and angle compensation commands for the six-axis robotic arm. The data storage module records historical deviation data.

9. The motor stator sheet angle positioning system according to claim 6, characterized in that: The six-axis robotic arm end effector includes: an electromagnetic chuck with adjustable suction force from 0 to 50 N; a six-dimensional force sensor with a range of ±200 N; and a miniature rotary motor with a resolution of 0.01°. It also includes a secondary calibration unit: Laser displacement sensor to measure the height difference at the edges of a metal sheet; Calibration algorithm: If the height difference Δh > 10 μm, then calculate the compensation angle. , where D is the diameter of the iron sheet.

10. The motor stator sheet angle positioning system according to claim 6, characterized in that: The industrial camera and processing unit communicate via the GigE Vision protocol; The processing unit is synchronized with the six-axis robotic arm via an EtherCAT bus; Real-time data transmission latency ≤1ms.

Images