Automatic calibration method for stator core lamination of brushless motor

Abstract

The invention relates to the technical field of motor stator lamination calibration, in particular to a brushless motor stator core lamination automatic calibration method, which comprises the following steps: S1, turntable dynamic balance calibration: scanning the working surface of a turntable through a laser interferometer to generate a three-dimensional morphology error graph; a plurality of groups of piezoelectric actuators are driven to compensate the planeness; step S2, iron sheet grabbing: spraying a silicon nitride coating on the surface of the iron sheet, applying ultrasonic vibration when the iron sheet is released, and adjusting the current of an electromagnetic chuck based on force feedback; s3, performing multi-stage visual positioning; step S4, lamination error prediction; the LSTM model outputs a compensation amount; according to a model updating mechanism, when the prediction error is larger than 0.02 mm, automatic incremental learning is carried out; s5, pressure closed-loop control is carried out; and S6, active vibration suppression. According to the invention, the precision and quality of lamination can be obviously improved, the lamination error is reduced, the performance and reliability of the iron core are ensured, and the stability and production efficiency of equipment are improved at the same time.

Description

Technical Field

[0001] The present invention relates to the technical field of motor stator lamination calibration, and in particular to an automatic calibration method for brushless motor stator core laminations. Background Art

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

[0003] Currently, existing motor stator core lamination technology faces numerous challenges. During the lamination process, lamination accuracy is difficult to control due to factors such as material properties, processing technology, and equipment precision. Material non-uniformity can lead to deviations in the laminations during stacking, impacting the core's overall performance. Difficulty controlling lamination accuracy can lead to a range of serious consequences. Insufficient lamination accuracy can lead to uneven magnetic field distribution within the motor, increasing losses and reducing efficiency. It can also cause vibration and noise in the motor, impacting its service life and reliability. Furthermore, low-precision laminations may require more manual intervention and subsequent processing, increasing production costs and cycle time. Therefore, improving motor stator core lamination technology and increasing lamination accuracy have become pressing challenges for the motor industry. Summary of the Invention

[0004] To solve the above problems, the present invention provides an automatic calibration method for brushless motor stator core laminations, which can significantly improve the accuracy and quality of laminations, reduce lamination errors, ensure the performance and reliability of the core, and at the same time improve the stability of the equipment and production efficiency.

[0005] The technical solution adopted by the present invention is: a method for automatically calibrating the stator core laminations of a brushless motor, comprising the following steps: Step S1, dynamic balance calibration of the turntable: Scanning the turntable working surface with a laser interferometer to generate a three-dimensional topography error map; driving multiple sets of piezoelectric actuators to compensate for flatness; Step S2, iron sheet grasping: spraying a silicon nitride coating on the surface of the iron sheet, applying ultrasonic vibration when releasing the iron sheet, and adjusting the current of the electromagnetic chuck based on force feedback; Step S3, multi-level visual positioning: coarse positioning: 2D camera identifies the iron sheet outline; fine positioning: 3D line scan camera reconstructs the mortise and tenon 3D coordinates; Step S4, stacking error prediction: Input parameters: historical 10-layer deviation, ambient temperature and humidity, mechanical vibration frequency; LSTM model outputs compensation amount; Model update mechanism: automatic incremental learning when the prediction error is greater than 0.02mm; Step S5, closed-loop pressure control: Install a 100-point / cm² pressure sensor matrix on the press head; generate a pressure distribution cloud map and calculate the regional standard deviation σ; if σ is greater than 0.15 MPa, adjust the pressure by region using a proportional valve; Step S6, active vibration suppression: collect the turntable vibration spectrum through the acceleration sensor; generate an anti-phase acoustic wave signal to drive the active shock absorber; control target: suppress the amplitude to less than or equal to 5μm.

[0006] A further improvement to the above scheme is that step S1 includes a thermal deformation compensation sub-step: real-time data collection of the eight temperature zones of the turntable. When the maximum temperature difference ΔT is greater than or equal to 1.5°C: calculation of thermal expansion, piezoelectric actuator compensation, liquid cooling system startup, and temperature equalization are performed in sequence until ΔT is less than 0.5°C.

[0007] A further improvement to the above solution is that the closed-loop control logic of the piezoelectric actuator compensation in step S1 is: The laser interferometer provides real-time feedback of the flatness error value ε; Electric actuator compensation calculation formula:

[0008] in, ; ; When ε is less than 0.5μm for three consecutive times, compensation is stopped.

[0009] A further improvement to the above solution is that in step S3, the visual positioning process includes the following steps: Step S3.1, image preprocessing: remove noise by median filtering; Step S3.2, feature extraction: segment the mortise and tenon region using a U-Net neural network; In step S3.3, the deviation is calculated using the following formula: Among them, the actual center coordinates are: ; Theoretical center coordinates: ; Offset:

[0010] Step S3.4, compensation decision: If Δd is greater than 0.05 mm, trigger the robot arm to fine-tune.

[0011] A further improvement to the above solution is the dynamic compensation mechanism for the deviation calculation in step S3.3: Actual center coordinates: Obtained through sub-pixel edge fitting algorithm; Theoretical center coordinates: Dynamic correction according to the number of lamination layers n: ; in, is the reference coordinate, It is the historical compensation amount output by S4.

[0012] A further improvement to the above scheme is that the U-Net network structure includes: encoder: 4 layers of convolution; decoder: 4 layers of deconvolution; jump connection: fusing shallow texture and deep semantic features.

[0013] A further improvement to the above solution is that the training process of the LSTM model in step S4 includes: collecting 100,000 sets of production data, offline pre-training, deployment to edge computing nodes, online updating every 24 hours, and triggering learning implementation based on abnormal data.

[0014] A further improvement to the above solution is that the proportional valve pressure regulation algorithm in step S5 is: Low pressure area compensation value:

[0015] High pressure area pressure reduction value:

[0016] in, is the average pressure.

[0017] A further improvement to the above solution is that the physical constraints for regional voltage regulation in step S5 are: Single voltage regulation range limit: Pressure balance judgment standard: When σ≤0.1MPa for 5 consecutive times, exit the pressure regulation mode; if the σ drop after pressure regulation is less than 10%, trigger the S1 turntable recalibration process.

[0018] A further improvement to the above solution is to further include step S7, calibration quality verification: every 50 pieces are measured using a laser tracker; the measurement includes concentricity detection and parallelism detection; Concentricity test is considered qualified if it is less than or equal to 0.05mm; parallelism test is considered qualified if it is less than or equal to 0.04mm per 100mm.

[0019] The beneficial effects of the present invention are: Compared to existing motor stator laminations, the present invention uses a laser interferometer to scan the turntable's working surface during the dynamic balancing calibration of the turntable and generate a three-dimensional topographic error map, accurately capturing the actual condition of the turntable's working surface. This high-precision detection method clearly reveals microscopic errors on the turntable's surface, providing an accurate data basis for subsequent calibration. By utilizing multiple sets of piezoelectric actuators to compensate for flatness, the turntable's planarity can be accurately adjusted in real time, effectively reducing lamination errors caused by turntable flatness issues. This not only improves the turntable's stability but also creates an excellent foundation for subsequent sheet metal grasping and lamination operations, ensuring the initial accuracy of the entire calibration process. During the sheet metal grasping step, a silicon nitride coating is sprayed on the sheet metal surface, enhancing its wear and corrosion resistance and extending its service life. Applying ultrasonic vibration during the release of the sheet metal effectively reduces the attraction between the sheet metal and the electromagnetic chuck, allowing for smoother release. Force feedback-based current regulation of the electromagnetic chuck allows precise control of its suction force based on the actual grasping situation, preventing failure or damage to the sheet metal due to excessive or insufficient suction force, thereby improving the success rate and accuracy of sheet metal grasping. The multi-stage visual positioning process combines coarse positioning with a 2D camera and fine positioning with a 3D line scan camera, enabling rapid and accurate determination of the sheet metal's position and posture. The 2D camera identifies the sheet metal's contour, quickly determining its approximate position within a wide range and providing a foundation for subsequent fine positioning. The 3D line scan camera reconstructs the three-dimensional coordinates of the mortise and tenon groove, achieving high-precision positioning and ensuring accurate placement of the sheet metal during the stacking process, effectively reducing stacking position errors. The stacking error prediction process uses an LSTM model to output compensation values ​​based on historical 10-layer deviations, ambient temperature and humidity, and mechanical vibration frequency. This model fully accounts for the impact of multiple factors on stacking error. This data-driven prediction approach allows for early prediction of stacking errors and timely compensation. The model update mechanism, which automatically performs incremental learning when the prediction error exceeds 0.02mm, enables the model to continuously adapt to new operating conditions and environmental changes, improving the accuracy of error prediction and the model's adaptability. The closed-loop pressure control system utilizes a 100-point / cm² pressure sensor matrix installed in the press head to accurately monitor the pressure distribution during the press-fitting process in real time. A pressure distribution cloud map is generated and the regional standard deviation σ is calculated, visually reflecting the uniformity of the pressure distribution. If σ exceeds 0.15MPa, a proportional valve is used to adjust the pressure in different regions, enabling precise adjustments based on actual pressure conditions. This ensures uniform pressure distribution during the lamination process and avoids core deformation or damage caused by uneven pressure. The active vibration suppression system uses an accelerometer to collect the turntable vibration spectrum, providing real-time monitoring of the turntable's vibration. This generates an anti-phase acoustic signal to drive the active shock absorber, effectively suppressing turntable vibration and reducing the amplitude to less than or equal to 5μm.It not only reduces the impact of vibration on lamination accuracy, but also improves the stability and reliability of the entire equipment and extends the service life of the equipment.

[0020] Through the synergistic effect of multiple links, the present invention can significantly improve the accuracy and quality of laminations, reduce lamination errors, ensure the performance and reliability of the iron core, and at the same time improve the stability and production efficiency of the equipment, and has important practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of the process of the automatic calibration method of the brushless motor stator core laminations of the present invention; Figure 2 1 is a flow chart of the automatic calibration method for the stator core laminations of a brushless motor according to the present invention, including step S7; Figure 3 The figure is a flow chart of the sub-pixel edge fitting algorithm used in the present invention. DETAILED DESCRIPTION

[0022] To facilitate understanding of the present invention, the present invention will be described more fully below with reference to the accompanying drawings. Preferred embodiments of the present invention are shown in the accompanying drawings. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and comprehensive understanding of the present disclosure.

[0023] It should be noted that when an element is referred to as being “fixed to” another element, it may be directly on the other element or there may be an intermediate element. When an element is referred to as being “connected to” another element, it may be directly connected to the other element or there may be an intermediate element.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention pertains. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention. Figures 1 to 3 As shown, in one embodiment of the present invention, a method for automatically calibrating stator core laminations of a brushless motor is provided, comprising the following steps: Step S1, dynamic balance calibration of the turntable: Scanning the turntable working surface with a laser interferometer to generate a three-dimensional topography error map; driving multiple sets of piezoelectric actuators to compensate for flatness; Step S2, iron sheet grasping: spraying a silicon nitride coating on the surface of the iron sheet, applying ultrasonic vibration when releasing the iron sheet, and adjusting the current of the electromagnetic chuck based on force feedback; Step S3, multi-level visual positioning: coarse positioning: 2D camera identifies the iron sheet outline; fine positioning: 3D line scan camera reconstructs the mortise and tenon 3D coordinates; Step S4, stacking error prediction: Input parameters: historical 10-layer deviation, ambient temperature and humidity, mechanical vibration frequency; LSTM model outputs compensation amount; Model update mechanism: automatic incremental learning when the prediction error is greater than 0.02mm; Step S5, closed-loop pressure control: Install a 100-point / cm² pressure sensor matrix on the press head; generate a pressure distribution cloud map and calculate the regional standard deviation σ; if σ is greater than 0.15 MPa, adjust the pressure by region using a proportional valve; Step S6, active vibration suppression: collect the turntable vibration spectrum through the acceleration sensor; generate an anti-phase acoustic wave signal to drive the active shock absorber; control target: suppress the amplitude to less than or equal to 5μm.

[0025] In the dynamic balance calibration link of the turntable, this embodiment uses a laser interferometer to scan the turntable working surface and generate a three-dimensional topographic error map, which can accurately obtain the actual condition of the turntable working surface. This high-precision detection method can clearly present the microscopic errors on the turntable surface, providing an accurate data basis for subsequent calibration. By using multiple sets of piezoelectric actuators to compensate for flatness, the planar state of the turntable can be adjusted in real time and accurately, effectively reducing the stacking error caused by the turntable flatness problem. Not only does it improve the stability of the turntable, but it also creates good basic conditions for the subsequent iron sheet grasping and stacking operations, ensuring the starting accuracy of the entire calibration process. During the iron sheet grasping step, a silicon nitride coating is sprayed on the surface of the iron sheet, which enhances the wear resistance and corrosion resistance of the iron sheet surface and extends the service life of the iron sheet. Applying ultrasonic vibration when releasing the iron sheet can effectively reduce the adsorption force between the iron sheet and the electromagnetic suction cup, making the release of the iron sheet smoother. Force feedback-based current regulation of the electromagnetic chuck allows precise control of its suction force based on the actual grasping situation, preventing failure or damage to the sheet metal due to excessive or insufficient suction force, thereby improving the success rate and accuracy of sheet metal grasping. The multi-stage visual positioning process combines coarse positioning with a 2D camera and fine positioning with a 3D line scan camera, enabling rapid and accurate determination of the sheet metal's position and posture. The 2D camera identifies the sheet metal's contour, quickly determining its approximate position within a wide range and providing a foundation for subsequent fine positioning. The 3D line scan camera reconstructs the three-dimensional coordinates of the mortise and tenon groove, achieving high-precision positioning and ensuring accurate placement of the sheet metal during the stacking process, effectively reducing stacking position errors. The stacking error prediction process uses an LSTM model to output compensation values ​​based on historical 10-layer deviations, ambient temperature and humidity, and mechanical vibration frequency. This model fully accounts for the impact of multiple factors on stacking error. This data-driven prediction approach allows for early prediction of stacking errors and timely compensation. The model update mechanism, which automatically incrementally learns when the prediction error exceeds 0.02mm, enables the model to continuously adapt to new operating conditions and environmental changes, improving the accuracy of error prediction and the model's adaptability. The closed-loop pressure control phase utilizes a 100-point / cm² pressure sensor matrix installed on the press head to accurately monitor the pressure distribution during the press-fit process in real time. A pressure distribution contour map is generated and the regional standard deviation σ is calculated, visually reflecting the uniformity of the pressure distribution. If σ exceeds 0.15MPa, a proportional valve adjusts the pressure in different regions, enabling precise adjustments based on actual pressure conditions. This ensures uniform pressure distribution during lamination and avoids core deformation or damage caused by uneven pressure. The active vibration suppression phase uses an accelerometer to collect the turntable's vibration spectrum, providing real-time monitoring of the turntable's vibration. This generates an anti-phase acoustic signal to drive the active damper, effectively suppressing turntable vibration, reducing the amplitude to less than or equal to 5μm. This not only reduces the impact of vibration on lamination accuracy but also improves the stability and reliability of the entire equipment, extending its service life.This embodiment can significantly improve the accuracy and quality of laminations, reduce lamination errors, ensure the performance and reliability of the core, and at the same time improve the stability and production efficiency of the equipment, and has important practical application value.

[0026] Step S1 includes a thermal deformation compensation sub-step: real-time data collection from the turntable's eight temperature zones. When the maximum temperature difference ΔT is greater than or equal to 1.5°C, the following steps are performed: thermal expansion calculation, piezoelectric actuator compensation, liquid cooling system activation, and temperature equalization to a ΔT of less than 0.5°C. In this embodiment, real-time data collection from the turntable's eight temperature zones enables comprehensive and accurate monitoring of the turntable's temperature conditions. By collecting data from all eight temperature zones, temperature variations in different areas of the turntable can be discerned, providing a reliable basis for subsequent accurate identification of thermal deformation issues. When the maximum temperature difference ΔT is greater than or equal to 1.5°C, a series of thermal deformation compensation operations are initiated. First, thermal expansion calculation is performed, which is the key to resolving thermal deformation issues. Accurate thermal expansion calculation clearly indicates the degree of deformation of the turntable caused by temperature differences, providing accurate quantitative guidance for subsequent compensation actions. Compensation is then performed using a piezoelectric actuator, which features fast response speed and high control precision. Based on the calculated thermal expansion, the turntable's flatness can be adjusted promptly, mitigating flatness errors caused by thermal deformation to a certain extent, ensuring the flatness of the turntable's working surface and creating an optimal foundation for subsequent lamination operations. The liquid cooling system is then activated, offering efficient heat dissipation. It quickly dissipates heat generated by the turntable's excessive temperature, accelerating the temperature equalization process. In cases of severe thermal deformation, the timely intervention of the liquid cooling system can effectively reduce the temperature and mitigate the effects of thermal expansion on the turntable. Ultimately, the temperature is equalized to a ΔT of less than 0.5°C, ensuring consistent temperatures across all areas of the turntable. This significantly reduces thermal deformation, ensures turntable stability and precision, and contributes to improved accuracy and reliability throughout the lamination process. A stable temperature environment reduces lamination errors caused by thermal deformation, improves the quality of the stator core laminations, and ultimately enhances the performance and reliability of the brushless motor.

[0027] The closed-loop control logic of the piezoelectric actuator compensation in step S1 is as follows: the laser interferometer provides real-time feedback of the flatness error value ε; Electric actuator compensation calculation formula:

[0028] in, ; ; When ε is less than 0.5μm for three consecutive times, compensation is stopped. In this embodiment, the core advantage of the closed-loop control logic lies in its high accuracy and real-time performance. The laser interferometer provides real-time feedback of the flatness error value ε, which enables the system to always grasp the actual flatness of the turntable working surface. This real-time monitoring allows the system to respond quickly according to the dynamic changes of the error, ensuring the timeliness and accuracy of the calibration process. The flatness error at each moment can be accurately captured, providing a reliable data basis for subsequent compensation operations. The setting of the calculation formula for the compensation amount of the piezoelectric actuator provides a scientific basis for the compensation operation. It takes into account a variety of factors, making the calculation of the compensation amount more reasonable and accurate. Determining the compensation amount based on the real-time feedback error value ε can ensure that the compensation force applied by the piezoelectric actuator is just right, and neither over-compensation will cause new errors to occur, nor under-compensation will fail to effectively solve the flatness problem.

[0029] See Figure 3 As shown, in step S3, the visual positioning process includes the following steps: Step S3.1, image preprocessing: remove noise by median filtering; Step S3.2, feature extraction: segment the mortise and tenon region using a U-Net neural network; In step S3.3, the deviation is calculated using the following formula: Among them, the actual center coordinates are: ; Theoretical center coordinates: ; Offset:

[0030] Step S3.4, compensation decision: If Δd is greater than 0.05 mm, trigger the robot arm to fine-tune.

[0031] In the above embodiment, median filtering is used to remove noise during the image preprocessing stage. During the image acquisition process, various interferences inevitably generate noise, which can seriously affect subsequent feature extraction and positioning. Median filtering effectively smooths the image, removes random noise, and preserves edge information. After median filtering, the image becomes clearer, laying a solid foundation for accurate feature extraction and improving the accuracy and reliability of the entire visual positioning process. During the feature extraction phase, a U-Net neural network is used to segment the mortise and tenon region. The U-Net neural network has powerful image segmentation capabilities and can accurately identify the mortise and tenon region. For brushless motor stator core laminations, accurate mortise and tenon identification is crucial, serving as a key feature for determining the position and posture of the iron sheets. The U-Net neural network can quickly and accurately segment the mortise and tenon region against a complex image background, significantly improving the efficiency and accuracy of feature extraction. Deviation calculation is performed using a clear formula, accurately calculating the offset between the actual center coordinates and the theoretical center coordinates. This precise calculation provides reliable data support for subsequent compensation decisions. Accurate deviation calculation provides a clear understanding of the positional deviation of the iron sheets, providing a basis for further adjustments. In the compensation decision-making process, when the offset Δd exceeds 0.05mm, the robotic arm triggers fine-tuning. This allows for timely detection of large deviations in the iron sheet position and prompt adjustments. Fine-tuning by the robotic arm ensures the iron sheet is accurately positioned, effectively reducing lamination errors. This precise compensation mechanism improves lamination accuracy and ensures the quality and performance of the brushless motor stator core.

[0032] Step S3.3 Dynamic compensation mechanism for deviation calculation: Actual center coordinates: Obtained through sub-pixel edge fitting algorithm; Theoretical center coordinates: Dynamic correction according to the number of lamination layers n: ; in, is the reference coordinate, It is the historical compensation amount output by S4.

[0033] In the above embodiment, a sub-pixel edge fitting algorithm is used to obtain the actual center coordinates, significantly improving coordinate acquisition accuracy. While traditional coordinate acquisition methods may only achieve pixel-level accuracy, the sub-pixel edge fitting algorithm can further enhance this accuracy to sub-pixel levels. This allows for more precise positioning of the actual position of the iron sheet, reducing stacking errors caused by inaccurate coordinate acquisition and providing a more reliable data basis for subsequent deviation calculations. The theoretical center coordinates are dynamically corrected based on the number of stacking layers, n. During the stacking process, as the number of layers increases, the theoretical position may be affected by various factors, such as cumulative errors and environmental changes, requiring corresponding adjustments. By combining the reference coordinates with the historical compensation output by S4, deviations from previous stacking processes can be fully accounted for. The historical compensation reflects adjustments made during previous stacking operations. By incorporating this historical compensation into the calculation of the theoretical center coordinates, the theoretical position can be dynamically adjusted based on the actual stacking conditions, ensuring that the theoretical center coordinates more closely align with actual stacking requirements. This dynamic compensation mechanism effectively reduces the impact of cumulative errors. During multi-layer stacking, cumulative errors can become increasingly amplified, significantly impacting the final stacking accuracy. By dynamically correcting the theoretical center coordinates, errors can be compensated and corrected in a timely manner, preventing error accumulation. Even with a large number of lamination layers, each layer can be accurately stacked, improving the accuracy and stability of the entire stator core lamination.

[0034] The above embodiment is implemented under the conditions of the following table:

[0035] The experimental results are shown in the following table:

[0036] The U-Net network architecture comprises an encoder with four convolutional layers, a decoder with four deconvolutional layers, and skip connections that fuse shallow texture features with deep semantic features. In this embodiment, the encoder employs four convolutional layers, enabling effective feature extraction and hierarchical feature construction from the input image. The convolution operation automatically learns local features within the image. As the number of convolutional layers increases, the network is able to capture increasingly abstract, high-level features. In the visual localization scenario of stator core laminations, four convolutional layers can extract key features such as grooves and edges from the image. These features are crucial for accurately identifying the position and posture of the stator core and provide a reliable basis for subsequent deviation calculation and calibration. Furthermore, multiple layers of convolution gradually reduce the size of the feature map, reducing data volume, improving computational efficiency, and speeding up the entire visual localization process. The decoder employs four deconvolutional layers, which upsample the feature map extracted by the encoder to a size similar to the original image. Deconvolution restores and refines the feature information compressed by the encoder, gradually recovering image details. During automatic calibration of stator core laminations, the decoder uses deconvolution to convert the feature information extracted by the encoder into specific image regions, accurately segmenting key areas such as the mortise and tenon, enabling the robotic arm to perform precise operations based on these segmentation results. Skip connections enable the U-Net network structure to fuse shallow texture and deep semantic features. Shallow texture features contain detailed image information, such as edges and textures, while deep semantic features contain the overall structure and semantic information of the image. By fusing these two features through skip connections, the advantages of each are fully utilized. During automatic calibration of stator core laminations, the fused features contain both detailed information and overall semantic information of the mortise and tenon, enabling more accurate identification and location of the mortise and tenon region, improving segmentation accuracy and robustness.

[0037] The LSTM model training process in step S4 includes: collecting 100,000 sets of production data, performing offline pre-training, deploying the model to edge computing nodes, and performing online updates every 24 hours. Learning is triggered by abnormal data. In this embodiment, collecting 100,000 sets of production data for offline pre-training provides the model with a rich and representative sample. This large amount of data enables the LSTM model to learn the complex patterns and patterns in the brushless motor stator core lamination process. In actual production, the lamination process is influenced by multiple factors, such as material properties, equipment status, and environmental conditions. These factors intertwine and create a complex production environment. By learning from 100,000 sets of data, the model can capture the relationships between these complex factors, accurately predicting and judging various conditions during the lamination process. Offline pre-training ensures that the model has a certain level of intelligence and accuracy before entering practical application, providing a solid foundation for subsequent automatic calibration. Deploying the trained model to edge computing nodes enables rapid data processing and real-time response. The proximity of edge computing nodes to the production site reduces data transmission latency. Real-time performance is crucial in the brushless motor stator core lamination process. If stacking deviations or other anomalies occur, prompt adjustments and calibration are required. Edge computing nodes can locally process and analyze data, making rapid decisions and adjusting calibration parameters in a timely manner, improving the efficiency and accuracy of automatic calibration. 24-hour online model updates enable the model to adapt to dynamic changes in the production process. The production environment is constantly changing, and factors such as material batch differences and equipment wear can cause fluctuations in production data. Through online updates, the model can continuously learn from new data and adjust its parameters to maintain accurate predictions and judgments about the production process. This dynamic adaptability enables the model to maintain efficient and accurate performance over the long term. Abnormal data triggers a real-time learning mechanism, further enhancing the model's robustness and adaptability. Abnormal data may indicate new conditions or problems in the production process. The model can promptly learn and analyze this abnormal data and adjust its decision-making strategy to address new challenges. For example, when encountering significant changes in material properties, the model can quickly adjust calibration parameters through real-time learning to ensure stacking quality is not affected.

[0038] The proportional valve pressure regulation algorithm in step S5 is as follows: Low pressure area compensation value:

[0039] High pressure area pressure reduction value:

[0040] in, is the average pressure.

[0041] In the above-mentioned embodiment, during the lamination process, the pressure conditions in different areas have a crucial impact on the quality of the laminations. The proportional valve's regional pressure regulation algorithm divides the pressure areas into low-pressure and high-pressure zones, and performs targeted pressure adjustments for each. For the low-pressure zone, pressure compensation is performed using a specific compensation value formula. During the actual lamination process, the low-pressure zone may experience insufficient pressure due to various factors, such as equipment structure and material properties. This insufficient pressure can result in a loose fit between the laminations, thereby affecting the overall performance of the stator core. By compensating the low-pressure zone, the pressure in this area is ensured to reach an appropriate level, ensuring a tight fit between the laminations and improving the quality and stability of the laminations. For example, after pressure compensation, the gaps between the laminations are reduced, effectively reducing vibration and noise during motor operation, thereby improving motor efficiency and reliability. For the high-pressure zone, pressure is adjusted using a pressure reduction value formula. Excessive pressure in the high-pressure zone can damage the laminations, such as causing deformation or cracking. Excessive pressure can also increase equipment wear and reduce its service life. By reducing the pressure in high-pressure areas, the pressure can be controlled within a safe and reasonable range, protecting the laminations and equipment from damage. This ensures the quality of the laminations and reduces equipment maintenance costs and replacement frequency. Using average pressure as a reference makes the pressure regulation process more scientific and rational. Average pressure reflects the overall pressure distribution. Using this as a basis for regional pressure regulation can better balance pressure differences between different areas and achieve uniform pressure distribution throughout the lamination process. This uniform pressure distribution helps improve the consistency and precision of the laminations, allowing each lamination to be accurately placed in its intended location, further enhancing the performance and quality of the brushless motor stator core.

[0042] Physical constraints for regional voltage regulation in step S5: Single voltage regulation amplitude limit: Pressure balance judgment standard: When σ≤0.1MPa for 5 consecutive times, exit the voltage regulation mode; if the σ decreases by less than 10% after voltage regulation, trigger the S1 turntable recalibration process. In this embodiment, the single voltage regulation amplitude limit can avoid the adverse effects of excessive voltage regulation on the lamination. During the lamination process of the stator core of a brushless motor, sudden and large changes in pressure may cause the lamination to deform, displace, or even be damaged. For example, if the single voltage regulation amplitude is too large, the position of the lamination that has been calibrated may be changed, destroying the previous calibration results. Limiting the single voltage regulation amplitude can make the pressure adjustment process smoother, gradually approaching the ideal pressure state, ensuring the stability and accuracy of the lamination during the voltage regulation process, and thus improving the lamination quality of the stator core. In the pressure balance judgment standard, when σ≤0.1MPa for 5 consecutive times, exit the voltage regulation mode. This setting ensures the effectiveness and accuracy of the voltage regulation process. σ represents pressure fluctuation. If it remains less than or equal to 0.1 MPa for five consecutive times, the pressure has reached a relatively balanced state. Exiting the pressure regulation mode at this point avoids over-regulation, saving time and energy while also preventing new pressure imbalances that may result from over-adjustment. This precise determination criterion helps improve pressure regulation efficiency and the overall lamination automatic calibration process. If the decrease in σ after pressure regulation is less than 10%, the S1 turntable recalibration process is triggered, enhancing the robustness of the calibration system. A small decrease in pressure fluctuation after pressure regulation indicates that the current pressure regulation operation may not have achieved the expected results, possibly due to the position of the turntable or other factors affecting the pressure distribution. Triggering the S1 turntable recalibration process allows the turntable to be recalibrated, troubleshooting the problem at its source and ensuring more effective subsequent pressure regulation operations. This feedback mechanism allows for timely detection and resolution of problems during the pressure regulation process, improving the success rate of lamination automatic calibration and ensuring the stable and reliable quality of brushless motor stator cores.

[0043] The calibration quality is also verified in step S7: every 50 sheets are measured using a laser tracker. This measurement includes both concentricity and parallelism testing. Concentricity less than or equal to 0.05 mm is considered acceptable, while parallelism less than or equal to 0.04 mm per 100 mm is considered acceptable. In this embodiment, every 50 sheets are measured using a laser tracker. This regular monitoring approach allows for timely detection of potential calibration issues. Since the production process can be affected by various factors, such as equipment wear and environmental changes, calibration results may deviate as the number of laminations increases. By measuring every 50 sheets, deviations can be detected and corrected before they accumulate and affect motor performance, ensuring the stability of the stator core lamination quality. The measurement includes both concentricity and parallelism testing, both of which are critical to the proper operation of brushless motors. A concentricity less than or equal to 0.05 mm is considered acceptable. Good concentricity ensures smooth rotation of the motor's rotor within the stator, reducing vibration and noise. Substandard concentricity can lead to additional friction and energy loss during motor operation, reducing the motor's efficiency and service life. Parallelism testing is considered acceptable if it is less than or equal to 0.04mm per 100mm. This qualified parallelism ensures uniform stress distribution between the stator core laminations, preventing deformation or damage due to excessive localized pressure. This helps improve the stability and reliability of the motor and reduces the probability of failure. Strict qualification standards provide a clear measure of calibration quality. Clearly defining the acceptable ranges for concentricity and parallelism provides quantifiable targets for the calibration process. Operators can evaluate and adjust calibration results based on these standards to ensure that each batch of stator core laminations meets high quality requirements.

[0044] The above embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A method for automatically calibrating a brushless motor stator core lamination, characterized by: The steps include: Step S1, dynamic balance calibration of the turntable: Scanning the turntable working surface with a laser interferometer to generate a three-dimensional topography error map; driving multiple sets of piezoelectric actuators to compensate for flatness; Step S2, iron sheet grasping: spraying a silicon nitride coating on the surface of the iron sheet, applying ultrasonic vibration when releasing the iron sheet, and adjusting the current of the electromagnetic chuck based on force feedback; Step S3, multi-level visual positioning: coarse positioning: 2D camera identifies the outline of the iron sheet; Precise positioning: 3D line scan camera reconstructs the 3D coordinates of the mortise and tenon; Step S4, stacking error prediction: Input parameters: historical 10-layer deviation, ambient temperature and humidity, mechanical vibration frequency; LSTM model outputs compensation amount; Model update mechanism: automatic incremental learning when the prediction error is greater than 0.02mm; Step S5, closed-loop pressure control: Install a 100-point / cm² pressure sensor matrix on the press head; generate a pressure distribution cloud map and calculate the regional standard deviation σ; if σ is greater than 0.15 MPa, adjust the pressure by region using a proportional valve; Step S6, active vibration suppression: collecting the turntable vibration spectrum through the acceleration sensor; Generate an anti-phase acoustic wave signal to drive the active shock absorber; control objective: suppress the amplitude to less than or equal to 5μm.

2. The automatic calibration method for the stator core laminations of a brushless motor according to claim 1, characterized in that: Step S1 includes the thermal deformation compensation sub-step: Real-time data collection of the eight temperature zones of the turntable. When the maximum temperature difference ΔT is greater than or equal to 1.5°C: The thermal expansion is calculated, the piezoelectric actuator is compensated, the liquid cooling system is started, and the temperature is balanced until ΔT is less than 0.5°C.

3. The automatic calibration method for the stator core laminations of a brushless motor according to claim 1, characterized in that: Closed-loop control logic for piezoelectric actuator compensation in step S1: The laser interferometer provides real-time feedback of the flatness error value ε; Electric actuator compensation calculation formula: in, ; ; When ε is less than 0.5μm for three consecutive times, compensation is stopped.

4. The automatic calibration method for the stator core laminations of a brushless motor according to claim 1, characterized in that: In step S3, the visual positioning process includes the following steps: Step S3.1, image preprocessing: remove noise by median filtering; Step S3.2, feature extraction: segment the mortise and tenon region using a U-Net neural network; In step S3.3, the deviation is calculated using the following formula: Among them, the actual center coordinates are: ; Theoretical center coordinates: ; Offset: Step S3.4, compensation decision: If Δd is greater than 0.05 mm, trigger the robot arm to fine-tune.

5. The automatic calibration method for the stator core laminations of a brushless motor according to claim 4, characterized in that: Step S3.3 Dynamic compensation mechanism for deviation calculation: Actual center coordinates: Obtained through sub-pixel edge fitting algorithm; Theoretical center coordinates: Dynamic correction according to the number of lamination layers n: ； in, is the reference coordinate, It is the historical compensation amount output by S4.

6. The automatic calibration method for the stator core laminations of a brushless motor according to claim 4, characterized in that: The U-Net network structure includes: encoder: 4 layers of convolution; Decoder: 4 layers of deconvolution; skip connection: fusion of shallow texture and deep semantic features.

7. The automatic calibration method for the stator core laminations of a brushless motor according to claim 1, characterized in that: The training process of the LSTM model in step S4 includes: collecting 100,000 sets of production data, offline pre-training, deployment to edge computing nodes, online updates every 24 hours, and triggering learning based on abnormal data.

8. The automatic calibration method for the stator core laminations of a brushless motor according to claim 1, characterized in that: The proportional valve pressure regulation algorithm in step S5 is as follows: Low pressure area compensation value: High pressure area pressure reduction value: in, is the average pressure.

9. The automatic calibration method for the stator core laminations of a brushless motor according to claim 8, characterized in that: Physical constraints for regional voltage regulation in step S5: Single voltage regulation range limit: Pressure balance judgment standard: When σ≤0.1MPa for 5 consecutive times, exit the pressure regulation mode; if the σ drop after pressure regulation is less than 10%, trigger the S1 turntable recalibration process.

10. The automatic calibration method for the stator core laminations of a brushless motor according to claim 1, characterized in that: The step S7 is also included, calibration quality verification: every 50 pieces are measured using a laser tracker; the measurement includes concentricity detection and parallelism detection; Concentricity test is considered qualified if it is less than or equal to 0.05mm; parallelism test is considered qualified if it is less than or equal to 0.04mm per 100mm.

Images