Magnetic field simulation and parameter optimization method and system for permanent magnet servo motor rotor

Abstract

The invention relates to the field of servo motor parameter optimization, in particular to a magnetic field simulation and parameter optimization method and system for a permanent magnet servo motor rotor. The method comprises the following steps: acquiring a rotor global magnetic field monitoring curve in the operation process of a servo motor rotor, and constructing a global magnetic field accurate compensation model; performing regional magnetic circuit structure analysis on the global magnetic field accurate compensation model, performing dynamic magnetic field logic mapping modeling, and constructing a real-time regional magnetic field logic map; performing magnetic field state change analysis on the global magnetic field accurate compensation model, performing multi-node change track correlation analysis, and constructing a multi-node magnetic field coupling dependency network; and performing short-term magnetic field evolution prediction based on the multi-node magnetic field coupling dependency network, and performing comprehensive electromagnetic performance evaluation to obtain an electromagnetic performance evaluation report. According to the method, the sudden change of the magnetic field is accurately and quickly identified, so that the magnetic field parameters are iteratively adjusted and optimized, and the service life and the performance of the servo motor are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of servo motor parameter optimization, and in particular to a magnetic field simulation and parameter optimization method and system for a permanent magnet servo motor rotor. BACKGROUND

[0002] With the rapid development of industrial automation and intelligent manufacturing, permanent magnet servo motors, as high-performance driving devices, are widely used in robots, numerical control machine tools, electric vehicles and precision instruments due to their high efficiency, rapid response and compact structure. Among the many types of servo motors, permanent magnet servo motors, with their high power density and excellent dynamic performance, have become the core driving force in modern industrial control systems. However, the magnetic field characteristics of the rotor of a permanent magnet servo motor directly affect the operating efficiency, control accuracy and reliability of the motor, so accurate simulation of the magnetic field and reasonable parameter optimization have become key technical problems to ensure the performance of the motor.

[0003] Traditional magnetic field simulation of permanent magnet servo motors mainly relies on numerical calculation methods such as finite element analysis (FEA), which can well simulate the static distribution of the motor magnetic field. However, in practical applications, due to the complexity of the rotor motion state, the nonlinear characteristics of the magnetic material and the existence of multi-physical field coupling effects, the simulation accuracy and efficiency are limited. In addition, existing parameter optimization is mainly based on experience or single-objective optimization, which is difficult to consider multiple factors such as electromagnetic performance, thermal effects and mechanical stability, resulting in optimization results that are difficult to achieve optimal performance requirements. On the other hand, with the advancement of sensing technology and real-time monitoring methods, it is possible to collect multi-dimensional magnetic field data during motor operation, and data-driven magnetic field analysis and dynamic optimization have gradually become a research hotspot. Using real-time magnetic field data for dynamic simulation and parameter adjustment can significantly improve the accuracy of the simulation model, and also enable rapid response and adjustment of rotor magnetic field anomalies, enhancing the robustness and adaptability of the motor. However, there is currently a lack of a comprehensive solution that integrates multi-modal data acquisition, real-time dynamic simulation and intelligent optimization and control for the magnetic field simulation and parameter optimization of permanent magnet servo motor rotors. Traditional methods are difficult to achieve efficient and accurate prediction and control in the face of complex magnetic field coupling, multi-node changes and transient mutations, limiting the further improvement and application expansion of motor performance. SUMMARY

[0004] To solve the above technical problems, a magnetic field simulation and parameter optimization method and system for a permanent magnet servo motor rotor are proposed to solve at least one of the above technical problems.

[0005] To achieve the above purpose, the present application provides a magnetic field simulation and parameter optimization method for a permanent magnet servo motor rotor, comprising the following steps: Step S1: Collect the rotor global magnetic field monitoring curve during the operation of the servo motor rotor, and perform real-time magnetic flux density gradient calculation and local magnetic field compensation modeling to construct a global magnetic field accurate compensation model; Step S2: Analyze the regional magnetic circuit structure of the global magnetic field accurate compensation model, and perform dynamic magnetic field logical mapping modeling to construct a real-time regional magnetic field logical atlas; Step S3: Analyze the magnetic field state change of the global magnetic field accurate compensation model, and perform multi-node change trajectory correlation analysis to construct a multi-node magnetic field coupling dependency network; Step S4: Based on the multi-node magnetic field coupling dependency network, perform short-term magnetic field evolution prediction, and perform comprehensive electromagnetic performance evaluation to obtain an electromagnetic performance evaluation report; Step S5: According to the real-time regional magnetic field logical atlas, perform real-time magnetic field transient mutation recognition, and perform space-time evolution tracking to obtain a magnetic field mutation diffusion distribution model; Step S6: According to the electromagnetic performance evaluation report and the magnetic field mutation diffusion distribution model, perform magnetic field mutation control iterative simulation, and perform intelligent parameter optimization planning to execute intelligent control optimization of the permanent magnet servo motor.

[0006] In the present specification, a magnetic field simulation and parameter optimization system for a permanent magnet servo motor rotor is provided for executing the magnetic field simulation and parameter optimization method for a permanent magnet servo motor rotor as described above, comprising: A magnetic field distribution calculation module is configured to collect the rotor global magnetic field monitoring curve during the operation of the servo motor rotor, and perform real-time magnetic flux density gradient calculation and local magnetic field compensation modeling to construct a global magnetic field accurate compensation model; A magnetic field logical analysis module is configured to analyze the regional magnetic circuit structure of the global magnetic field accurate compensation model, and perform dynamic magnetic field logical mapping modeling to construct a real-time regional magnetic field logical atlas; A trajectory correlation module is configured to analyze the magnetic field state change of the global magnetic field accurate compensation model, and perform multi-node change trajectory correlation analysis to construct a multi-node magnetic field coupling dependency network; A magnetic field evolution prediction module is configured to perform short-term magnetic field evolution prediction based on the multi-node magnetic field coupling dependency network, and perform comprehensive electromagnetic performance evaluation to obtain an electromagnetic performance evaluation report; A mutation diffusion distribution module is configured to perform real-time magnetic field transient mutation recognition according to the real-time regional magnetic field logical atlas, and perform space-time evolution tracking to obtain a magnetic field mutation diffusion distribution model; A magnetic field parameter optimization module is configured to perform magnetic field mutation control iterative simulation according to the electromagnetic performance evaluation report and the magnetic field mutation diffusion distribution model, and perform intelligent parameter optimization planning to execute intelligent control optimization of the permanent magnet servo motor.

[0007] The beneficial effects of the present application are as follows: the high-precision magnetic field sensor collects the global magnetic field change data during the operation of the rotor, providing the original basis for subsequent modeling. Real-time calculation of the magnetic flux density gradient can timely reflect the local imbalance of the magnetic field distribution, thereby effectively identifying potential magnetic flux concentration areas or weakened areas. The introduction of local magnetic field compensation modeling mechanism improves the spatial resolution and local error correction capability of the global magnetic field model, enhances the dynamic response capability and accuracy of the model. Mapping the magnetic field changes to the rotor region structure helps to find the specific impact of structure design on the magnetic field distribution, and promotes structure optimization. The construction of regional magnetic field logical atlas enables the motor system to logically infer the dynamic relationship of the magnetic field in each region, improving the cognitive level of the magnetic field evolution. The atlas model has real-time updating capability and can be used as a dynamic reference for system control, realizing more responsive and rapid magnetic field control decision support. The magnetic field state change analysis realizes the leap from static magnetic field modeling to dynamic magnetic field behavior modeling, and can identify key transition points. By constructing a magnetic field coupling dependency network, the dynamic interaction relationship between each magnetic field node is revealed, which helps to master the flux linkage mode under complex systems. This network model provides a solid foundation for subsequent anomaly identification, early warning and fault pre-control, and has the ability to identify abnormal magnetic field linkage phenomena in advance. Through the coupling network, short-term magnetic field evolution prediction can be realized, which can predict magnetic field disorder or local anomalies in advance to support proactive regulation or maintenance planning. The evaluation model considers key parameters such as magnetic flux distribution, magnetic saturation, loss, and torque fluctuation, forming a systematic performance analysis mechanism. The electromagnetic performance evaluation report provides quantitative index support for parameter optimization, structure adjustment and control strategy formulation, enhancing the pertinence and scientificity of optimization. The mutation diffusion distribution model can effectively track the magnetic disturbance affected area, providing path prediction support for local magnetic disturbance isolation and intervention. Identifying mutation diffusion patterns can be used to optimize the robustness of the magnetic field structure and improve the anti-disturbance ability of the motor under extreme working conditions. Through the iterative process of regulation simulation and parameter optimization, a complete closed loop from model-evaluation-optimization-feedback is realized. Intelligent optimization not only improves the efficiency of the magnetic field, but also effectively reduces heat and loss, prolonging the service life of the motor. It has an intelligent planning interface integrated with existing motor control systems (such as DSP, FPGA controllers), promoting the automation and intelligentization of the system. BRIEF DESCRIPTION OF DRAWINGS

[0008] Figure 1 Figure 1 is a schematic flow chart of the steps of the method for magnetic field simulation and parameter optimization of a permanent magnet servo motor rotor according to the present application; Figure 2 Figure 2 is a detailed implementation step flow chart of step S1 according to the present application; Figure 3 Figure 3 is a detailed implementation step flow chart of step S2 according to the present application; Figure 4For the detailed implementation step flowchart of step S3. DETAILED DESCRIPTION

[0009] It should be understood that the specific embodiments described herein are merely illustrative of the present application and do not limit the present application.

[0010] The application example provides a magnetic field simulation and parameter optimization method and system for a permanent magnet servo motor rotor.

[0011] Please refer to Figures 1 to 4 The application provides a magnetic field simulation and parameter optimization method for a permanent magnet servo motor rotor, which comprises the following steps: Step S1: Collecting the rotor global magnetic field monitoring curve in the running process of the servo motor rotor, and performing real-time magnetic flux density gradient calculation and local magnetic field compensation modeling to construct a global magnetic field accurate compensation model; Step S2: Analyzing the regional magnetic circuit structure of the global magnetic field accurate compensation model, and performing dynamic magnetic field logical mapping modeling to construct a real-time regional magnetic field logical atlas; Step S3: Analyzing the magnetic field state change of the global magnetic field accurate compensation model, and performing multi-node change trajectory correlation analysis to construct a multi-node magnetic field coupling dependency network; Step S4: Based on the multi-node magnetic field coupling dependency network, performing short-term magnetic field evolution prediction, and performing comprehensive electromagnetic performance evaluation to obtain an electromagnetic performance evaluation report; Step S5: According to the real-time regional magnetic field logical atlas, performing real-time magnetic field transient mutation identification, and performing space-time evolution tracking to obtain a magnetic field mutation diffusion distribution model; Step S6: According to the electromagnetic performance evaluation report and the magnetic field mutation diffusion distribution model, performing magnetic field mutation regulation and control iterative simulation, and performing intelligent parameter optimization planning to execute permanent magnet servo motor intelligent regulation and control optimization operation.

[0012] In the embodiment of the application, please refer to Figure 1 The application provides a magnetic field simulation and parameter optimization method for a permanent magnet servo motor rotor, which comprises the following steps: Step S1: Collecting the rotor global magnetic field monitoring curve in the running process of the servo motor rotor, and performing real-time magnetic flux density gradient calculation and local magnetic field compensation modeling to construct a global magnetic field accurate compensation model; In this embodiment, the motor rotor is sampled in three-dimensional full-field under the rated operating conditions (such as 3000 rpm, temperature rise to 65°C) by a high-sensitivity three-dimensional Hall array magnetic sensor system (sampling accuracy up to ±0.5 mT, spatial resolution less than 1 mm). The sensor records the change curve of the magnetic flux density in the rotor outside and the air gap area with the rotation angle and time at a frequency of 1000 points per second, forming a magnetic field time sequence data sequence for one revolution of the rotor. Through the spatio-temporal joint analysis algorithm, the real-time magnetic flux density gradient calculation is carried out on the collected monitoring curve. This process takes the rotor rotation angle as the reference axis, combines the change rate of the magnetic flux density value in the radial and axial directions, and constructs a magnetic flux density gradient tensor field for identifying gradient mutation areas. These areas often reflect the uneven diffusion of magnetic flux, magnetic leakage or material saturation problems. For example, in the experiment, when the speed exceeds 3500 rpm, the air gap magnetic flux gradient abnormally rises (up to 1.8 mT / mm) between the rotation angles of 120° to 150°, corresponding to the existence of structural deviation or local material magnetic performance degradation phenomenon at the extreme edge of the rotor. A local magnetic field compensation model is constructed for these local abnormal areas. The model uses finite element simulation (such as ANSYS Maxwell) combined with experimental data fitting method to difference fit the actual magnetic flux density curve with the ideal symmetrical distribution model, generating a compensation factor matrix. The compensation model considers the magnetic flux density spatial distribution, operating conditions (current, temperature, speed) and material nonlinear permeability characteristics, and introduces dynamic correction coefficients to adapt to different load conditions. For example, corresponding to the gradient sudden increase area, the local compensation model suggests adding additional magnetic path or locally strengthening the pole shape within 0.3 mm of the pole edge to balance the magnetic field. Based on the local compensation model of the whole rotor area, through spatial fusion and gradient smoothing processing, a complete full-field accurate compensation model is constructed. This model can predict the magnetic field distribution deviation in real time under different operating conditions, and provide quantitative basis for subsequent parameter optimization and magnetic field control strategy. In the verification, after the compensation model is applied to the running simulation system, the magnetic field unevenness decreases by more than 25%, the gradient mutation rate decreases by 35%, the magnetic field distribution tends to be axisymmetric, significantly improving the electromagnetic performance stability and efficiency of the motor. This step constitutes the key starting point of the magnetic field simulation and intelligent parameter optimization process.

[0013] Step S2: Analyze the regional magnetic circuit structure of the full-field accurate compensation model, and perform dynamic magnetic field logical mapping modeling to construct a real-time regional magnetic field logical atlas; In this embodiment, the global magnetic field compensation model constructed from step S1 is used to perform multi-scale magnetic circuit slicing processing on the rotor magnetic field data. Taking the rotor shaft center as the origin, the inter-pole region is divided by every 15° angle, and the magnetic field data is sampled in the radial and axial directions respectively, and the magnetic induction intensity vector distribution of different slice planes is extracted. Each slice region is further subjected to magnetic field distribution derivation and magnetic field line fitting to identify the main closed magnetic circuit path. For example, in actual testing, the magnetic flux closed path in the pole edge region appears a local magnetic circuit fracture trend at 90°-105° and 255°-270°, which corresponds to the saturation of the core edge material and the uneven structure. Topological mapping and semantic deconstruction are performed on these identified magnetic circuit structures to construct a regional magnetic circuit topological network graph. The topological graph takes the pole center, pole boundary, and pole arc junction point as the core node, takes the magnetic field line direction as the directed edge, and labels the magnetic flux density value, permeability, path length, and other attributes of each path. Subsequently, combined with the magnetic flux density time series data in the running process, the dynamic change trend of the magnetic flux in the path is analyzed through the time sliding window mechanism to establish a "magnetic field dynamic response function" reflecting the dynamic evolution characteristics of the magnetic flux with speed, current, and temperature rise. For example, when the temperature rises to 75°C, the inductance of the magnetic flux path corresponding to a certain magnetic pole decreases by 10%, causing the magnetic flux response to lag about 0.2ms, and the path logic response label in the atlas is automatically updated. The topological structure and dynamic evolution function are fused to generate a "real-time regional magnetic field logic atlas". This atlas not only has a static topological structure, but also embeds time sequence magnetic flux state logic information, supporting the modeling of the logical cause-effect relationship of magnetic field evolution. In order to enhance the real-time response capability, the logic atlas also integrates an abnormal path recognition mechanism based on time series decision tree, which can automatically track the affected path and node when the magnetic flux abnormally deviates (such as ±20% or more). In experimental verification, the atlas is loaded into an online magnetic field monitoring system, which can real-time judge the magnetic flux response lag point and local overload region under speed fluctuation, and predict the sudden magnetic flux congestion trend, with a warning accuracy rate of more than 92%.

[0014] Step S3: Perform magnetic field state change analysis on the global magnetic field precise compensation model, and perform multi-node change trajectory correlation analysis to construct a multi-node magnetic field coupling dependency network; In this embodiment, key magnetic field nodes are selected from the precise compensation model as the analysis target, including: magnetic pole center point, pole boundary point, air gap center point, stator-rotor interface, and rotor axial symmetry point, etc., totaling about 50-100 monitoring points. Using three-dimensional magnetic field scanning data (spatial resolution 0.5 mm, sampling frequency 10 kHz), the magnetic flux density, magnetic induction intensity and direction of each node in a cycle of motor operation are tracked in time sequence, and the magnetic field state change trajectory curve of each node is obtained. Through time sequence analysis of the magnetic field state change curve, characteristic indexes such as change amplitude, response lag, frequency component and fluctuation amplitude are extracted. On this basis, the "magnetic field state change vector" of each node is constructed. It is found in the experiment that when the motor runs at 4000 rpm, the magnetic field fluctuation amplitude of the pole edge point can reach ±3.2 mT, and the response delay of the magnetic pole center point is 0.18 ms on average. These parameters reflect the spatiotemporal non-uniformity characteristics in the magnetic field dynamic process. Using multivariate dynamic time warping (DTW) algorithm, similarity matching and time sequence alignment analysis of the magnetic field state change vectors between all nodes are carried out, and a group of magnetic nodes with strong dynamic correlation are identified. For example, the magnetic field fluctuation trend between the pole arc transition zone and the stator tooth top is highly consistent, and they show obvious resonance coupling phenomenon under high frequency load. After obtaining the similarity matrix between nodes, a multi-node magnetic field coupling dependency network is further constructed. Each node in the network represents a spatial magnetic node, and the weight of the edge represents the coupling strength of the magnetic flux change between adjacent nodes. The weight calculation not only considers the fluctuation synchronization degree (such as Pearson correlation coefficient > 0.85), but also introduces the frequency energy coupling degree and response time lag factor. For example, under high load state, the coupling path between some nodes will be significantly enhanced, forming a "magnetic flux linkage block", which is the core area of efficient operation of the motor. The multi-node magnetic field coupling dependency network is constructed as a dynamic directed graph, which not only describes the coupling relationship between the magnetic field space nodes, but also can update the state according to the real-time running data, and predict the potential chain reaction of magnetic flux abnormality in a certain area to other areas. In practical application, the network is used for dynamic magnetic field stability evaluation and adaptive control system scheduling optimization, which can improve the magnetic field uniformity under high load operation, and the response time of the regulation and control is improved by more than 18% compared with the traditional control scheme, and the magnetic saturation feedforward prediction accuracy is improved to 93%.

[0015] Step S4: short-term magnetic field evolution prediction based on the multi-node magnetic field coupling dependency network, and comprehensive electromagnetic performance evaluation, to obtain an electromagnetic performance evaluation report; In this embodiment, the real-time magnetic field state data of the motor in operation is input into the network, and a short-term magnetic field evolution prediction model is established by using the node coupling weight and dynamic connection relationship in the network structure. The specific method is to use the long short-term memory network (LSTM) structure in the recurrent neural network (RNN), which can effectively capture the time sequence dependence and nonlinear coupling characteristics between multiple nodes. In the training stage, the collected magnetic field data is used, the time window is set to 100 ms, and the time step is 1 ms, so as to ensure that the model can sensitively capture the rapid changes and trends of the magnetic field. In the prediction process, the system relies on the coupling network structure to generate a short-term magnetic field evolution distribution map within 10-50 ms by iteratively calculating the future magnetic flux change of the node state. This map accurately reflects the evolution trend of the magnetic flux density and the magnetic field strength at different spatial nodes, especially at the rotor pole arc boundary and the air gap position. The predicted magnetic field fluctuation amplitude error is controlled within ±0.15 mT, meeting the high-precision monitoring requirements. Combined with the short-term predicted magnetic field distribution data, the comprehensive electromagnetic performance evaluation is carried out. This part mainly uses finite element analysis (FEA) technology to simulate and calculate the predicted magnetic field, focusing on key electromagnetic parameters such as magnetic flux density distribution uniformity, magnetic saturation risk, magnetic hysteresis loss, and eddy current loss. In the experimental setup, a spatial grid division of 0.1 mm and a time step of 0.5 ms are used for dynamic simulation, and the calculation results are used to evaluate the overall efficiency of the rotor and the potential thermal load distribution. In combination with the time sequence distribution fluctuation characteristics of the predicted magnetic field, statistical methods are used to analyze the fluctuation of the magnetic field strength, extract the peak frequency, fluctuation amplitude, and abnormal mutation indicators, and further evaluate the stability and electromagnetic compatibility of the motor in the short-term operation. By comparing different operating conditions (such as rated load and overload state), the influence of magnetic field changes on motor performance is determined, and the sensitive areas with large magnetic flux fluctuations are identified, providing guidance for subsequent local magnetic field compensation strategies. Based on the above analysis results, a complete electromagnetic performance evaluation report is formed. The report content covers short-term magnetic field prediction accuracy, key area electromagnetic indicators, potential failure risk points and their spatial distribution, as well as recommended optimization direction. The report also includes multi-dimensional charts and visual curves, which facilitate engineers to intuitively understand and make decisions on magnetic field dynamic changes and electromagnetic performance.

[0016] Step S5: Real-time magnetic field transient mutation recognition is performed according to the real-time regional magnetic field logical map, and a magnetic field mutation diffusion distribution model is obtained by tracking the space-time evolution; In this embodiment, a real-time regional magnetic field logical atlas is used as the basic data framework, which describes in detail the magnetic field distribution law, magnetic flux path and coupling relationship between nodes in different regions of the rotor. Through the input of real-time collected magnetic field data into the atlas, the system uses a time sequence anomaly detection algorithm to identify magnetic field transient mutations. The specific method uses an adaptive threshold detection algorithm based on a sliding window, with a window length of 20 ms and a step of 1 ms. This algorithm combines statistical characteristics (such as mean, variance) and transient signal energy mutation indicators to quickly and accurately capture sudden abnormal changes in the magnetic field. After the mutation point is identified, the system further locates the source of the transient mutation. Using the topological structure of the nodes in the magnetic field logical atlas, combined with a weighted graph search algorithm, the starting area of the mutation signal is determined, with a positioning accuracy of millimeter level (about 0.5 mm) in the rotor structure, thus marking the specific local magnetic field transient mutation area. For the mutation area, a spatio-temporal data fusion method is used to dynamically track the diffusion process of the magnetic field mutation. Specifically, multi-dimensional spatio-temporal filtering technology (such as Kalman filtering combined with spatial convolution filtering) is used to estimate the diffusion path of the mutation signal in space and time, generating a mutation diffusion evolution trajectory. In the experiment, the time resolution is set to 1 ms and the spatial resolution is set to 0.1 mm³, ensuring that subtle mutation diffusion dynamics can be captured. Based on these spatio-temporal trajectories, a magnetic field mutation diffusion distribution model is constructed. The model uses a numerical simulation method based on diffusion equations, combined with actual magnetic field physical properties (such as magnetic permeability, saturation characteristics) and rotor material parameters, to simulate how mutation energy diffuses within the rotor magnetic circuit. The diffusion coefficient in the model is calibrated through historical experimental data, usually with a value range of 0.05~0.15 m² / s. At the same time, the model takes into account the local nonlinear response of the magnetic field and the influence of multi-node coupling, making the simulation results closer to the real physical process. The model results are rendered in three dimensions for visualization, generating a dynamic demonstration atlas of magnetic field mutation diffusion distribution. This visualization not only shows the mutation starting point and diffusion path, but also reflects the spatial distribution of magnetic field strength over time, helping engineers intuitively understand the propagation mechanism and potential impact range of magnetic field transient mutations.

[0017] Step S6: Perform magnetic field mutation regulation iteration simulation based on the electromagnetic performance evaluation report and the magnetic field mutation diffusion distribution model, and perform intelligent parameter optimization planning to execute the intelligent regulation and optimization of the permanent magnet servo motor.

[0018] In this embodiment, according to the key information such as magnetic flux density distribution, magnetic field strength fluctuation and coupling characteristics provided by the electromagnetic performance evaluation report, combined with the mutation starting point, diffusion path and time evolution law identified in the magnetic field mutation diffusion distribution model, a simulation framework of magnetic field mutation regulation is constructed. The simulation framework based on finite element method (FEM) and multi-physical field coupling technology can truly simulate the evolution dynamics of magnetic field under different regulation parameters. The experimental parameter settings include magnetic permeability, rotor geometric parameters and material saturation curve, and the grid element size is controlled at the order of 0.1 millimeter to ensure the balance between simulation accuracy and calculation efficiency. The regulation iteration simulation is carried out. The regulation variables include excitation current, rotor magnetic pole gap fine tuning, local magnetic circuit compensation parameters, etc. The system adjusts the parameters to suppress the diffusion trend of magnetic field mutation in each iteration through the designed iteration algorithm (such as genetic algorithm combined with gradient descent optimization), and gradually approaches the optimal magnetic field distribution state. The length of each simulation iteration is about 5 minutes, and the number of iterations is set according to the convergence condition, generally controlled within 30 times to ensure the balance between calculation time and optimization effect. After each simulation ends, the system automatically extracts key performance indicators such as magnetic flux density uniformity, magnetic field peak value, coupling stability and electromagnetic loss, and compares them with the preset performance targets to form performance difference feedback. Based on these feedbacks, the intelligent parameter optimization module adjusts the next parameter search strategy using reinforcement learning algorithm in machine learning, gradually optimizes the regulation path, avoids falling into local optimum, and improves the overall regulation efficiency. The intelligent parameter optimization planning stage generates the final intelligent regulation scheme according to the regulation iteration simulation results, considering the magnetic field regulation effect, energy consumption and hardware realizability. The scheme includes specific excitation current adjustment curve, magnetic pole fine tuning amount and compensation equipment action instruction, and is output in the form of dynamic parameter configuration file for easy online execution and real-time update. The system also generates a visual report of the regulation effect, showing the change of magnetic field before and after regulation and the mutation suppression situation, which helps engineers to intuitively evaluate the optimization effect. Based on the intelligent regulation scheme, the intelligent regulation and optimization operation of the permanent magnet servo motor is carried out. The operation applies the optimized parameters to the actual rotor operation through the online control system to realize dynamic magnetic field adjustment, significantly reducing the frequency and intensity of magnetic field transient mutation. In the experimental environment, the magnetic field transient peak value after regulation is reduced by about 20%, the magnetic flux uniformity is improved by 15%, and the motor operating efficiency is improved by about 3%, verifying the effectiveness of the regulation method.

[0019] In this embodiment, referring to Figure 2 The detailed implementation steps of step S1 include: Detect the running state of the servo motor rotor, perform three-dimensional space magnetic field scanning, and obtain the rotor global magnetic field monitoring curve; Adaptive magnetic interference filtering optimization is performed on the rotor global magnetic field monitoring curve to obtain an adaptive filtering optimized monitoring curve; The adaptive filter optimization monitoring curve is analyzed in three-dimensional space distribution to obtain a rotor global magnetic field distribution map; Real-time magnetic flux density gradient calculation is performed based on the rotor global magnetic field distribution map to generate a magnetic flux density gradient map; The magnetic flux density gradient map is processed by spatial frequency domain transformation to construct a magnetic field frequency domain characteristic spectrum; The magnetic field frequency domain characteristic spectrum is decomposed layer by layer, and interlayer magnetic flux coupling calculation is performed to obtain multi-layer magnetic flux coupling parameters; Interlayer coupling difference analysis is performed according to the multi-layer magnetic flux coupling parameters, and local magnetic field compensation modeling is performed to construct a global magnetic field accurate compensation model.

[0020] In this embodiment, the dynamic three-dimensional magnetic field scanning of the permanent magnet servo motor rotor in the running state is carried out by a high-precision magnetic sensor array, and the spatial magnetic flux distribution is obtained in real time. The sensor type used is usually a three-axis Hall effect sensor, with a response frequency greater than 10 kHz, a measurement accuracy of ±0.1 mT, and a sensor layout density of one set per 20° inter-pole. A 360° full coverage is achieved within the entire rotation period. During the sampling process, the motor runs at a steady speed of 3000 rpm, and the data acquisition control system collects signals at a sampling frequency of 2 kHz to ensure that the time domain resolution meets the requirements of capturing dynamic magnetic field changes. The rotor angle position and magnetic field sampling value are synchronized through the position encoder to realize the reconstruction of the magnetic field data with consistent time and space. The collected data are processed by a spatial interpolation algorithm (such as three-dimensional spline interpolation) to form continuous rotor three-dimensional magnetic field monitoring curves, providing basic data support for subsequent magnetic field distribution analysis and modeling. Due to the influence of various interference factors (such as coil electromagnetic leakage, power fluctuations, and disturbance magnetic flux caused by mechanical vibration) during motor operation, the original monitoring curve often has noise and artifacts, which need to be filtered and optimized. The adaptive wavelet threshold filtering method is used to denoise the global magnetic field monitoring curve. The specific method is as follows: first, perform discrete wavelet transform (such as using db4 mother wavelet) to decompose the original curve into high and low frequency signals at multiple scales; then select an adaptive soft threshold function according to the local variance estimation to suppress high frequency noise; finally, reconstruct the signal to obtain a smooth and edge feature preserved magnetic field monitoring curve. Experiments show that this method can effectively suppress interference signals with a frequency higher than 500 Hz, while maintaining the magnetic flux change trend in the main frequency band (50-300 Hz), and the signal-to-noise ratio is improved by about 23%. The optimized curve more truly reflects the actual rotor magnetic field state and is a reliable basis for three-dimensional magnetic field reconstruction. The filtered monitoring curve is mapped back to the three-dimensional space to construct the rotor global magnetic field distribution map. By combining the sensor spatial layout matrix and the motor geometric parameters (such as rotor radius, pole pitch, air gap width, etc.), the inversion reconstruction algorithm is used to convert the magnetic flux value at each time node to the magnetic flux density distribution in the polar direction and the radial and axial directions. The spatial distribution map is expressed in the form of a vector field, containing the magnetic flux size and direction of each point. The three-dimensional vector visualization tool in Matlab or Python is used to draw magnetic field lines, which can clearly show the closed path and asymmetric distribution of the rotor magnetic field. In the experiment, the reconstruction accuracy error is controlled within ±2%, and the change trend of the magnetic field distribution under different load and speed conditions can be dynamically observed, revealing key phenomena such as magnetic saturation and leakage.

[0021] The magnetic flux density gradient map is used to reveal the spatial variation rate of the magnetic field change intensity and direction, and is an important tool for analyzing the non-uniformity of the magnetic field. In this step, the finite difference method or the finite element post-processing algorithm is used to calculate the gradient of the magnetic field vector distribution data. Based on the magnetic flux density vector B(x, y, z), the first-order partial derivatives in three orthogonal directions are calculated to form the magnetic flux density gradient tensor field. The gradient map can reveal the magnetic field mutation characteristics in the air gap area, the pole edge and the corner of the rotor core. In the experimental setup, the calculation step is controlled within 0.2 mm to ensure high-resolution gradient output. The generated gradient map is visualized using pseudo-color, with red representing the strong gradient area and blue representing the weak gradient area. This map provides key evidence for analyzing the magnetic field distortion source, identifying potential saturation area and edge demagnetization risk. The periodic characteristics and local distortion frequency components of the magnetic field are extracted, and the 3D-FFT processing is performed on the magnetic flux density gradient map to construct the spatial frequency domain feature spectrum of the magnetic field. By expanding the magnetic flux gradient in each direction in the frequency domain, the energy distribution spectrum of each spatial frequency component is obtained. The high-amplitude frequency band in the frequency spectrum corresponds to the periodic magnetic distortion, such as the slot effect and winding harmonics; the low-frequency component reflects the macro trend of the magnetic field. In the experiment, the frequency domain resolution is controlled at 0.5 Hz, and the Hamming window is used to suppress the sidelobe leakage. The constructed frequency domain feature spectrum reveals the causes of the non-ideal disturbance in the rotor magnetic field, and provides frequency domain basis for improving the magnetic circuit structure (such as adjusting the pole arc ratio) in the design end.

[0022] The layer-by-layer decomposition of the magnetic field frequency domain feature spectrum adopts wavelet packet decomposition technology to perform multi-resolution decomposition on the frequency domain signal and obtain the magnetic flux energy distribution at each frequency band level. By dividing the frequency spectrum into multiple sub-bands of low, medium and high frequencies, the degree of magnetic flux coupling between each layer is quantified. The coupling calculation uses cross-correlation coefficient and energy superposition ratio as the main indicators to analyze whether there is significant phase or energy coupling phenomenon between different frequency bands. It is found that there is cross-excitation coupling phenomenon between the high frequency layer (>1 kHz) and the medium frequency layer (300-800 Hz) of the rotor, and the cross-correlation coefficient reaches more than 0.65, indicating that the structure resonance or electromagnetic interference is affected. The coupling parameters are finally used to identify the conduction path and dissipation mechanism of the magnetic field energy between different structure layers. Based on the foregoing coupling parameters, the differences in magnetic flux transmission between each layer are analyzed, and the areas where the magnetic field imbalance or magnetic flux "short circuit" phenomenon is more serious are identified. Combined with the position and frequency response characteristics of the coupling difference area, a local magnetic field compensation model is constructed. The compensation model uses the parametric magnetic guide adjustment method to numerically adjust the magnetic permeability, pole shape size or magnetic performance of the material in the compensation area to simulate its compensation effect on the magnetic flux distribution. Then the compensation model is embedded in the overall three-dimensional magnetic field simulation system for iterative verification, and adjusted to minimize the coupling difference. By adjusting the magnetic permeability within ±5% range through the local compensation model, the magnetic field uniformity in the gradient mutation area can be improved by more than 20%, greatly optimizing the magnetic field consistency. Finally, a set of precise compensation model of global magnetic field is formed, which provides simulation reference for the structural optimization design of servo motor.

[0023] In the embodiment, the specific steps of adaptively filtering and optimizing the rotor global magnetic field monitoring curve to obtain an adaptively filtered and optimized monitoring curve are as follows: The rotor global magnetic field monitoring curve is decomposed into multiple time periods, and the magnetic field monitoring curves of the multiple time periods are extracted. The magnetic field monitoring curve is used to identify environmental magnetic interference, and the magnetic interference fluctuation time points are marked. The environmental magnetic interference strength is calculated according to the magnetic interference fluctuation time points, and the multi-time period interference distribution is analyzed to obtain a multi-time period interference distribution graph. The multi-time period interference distribution graph is used to identify the structural rules of environmental magnetic interference, and the structural rules of magnetic interference are obtained. Based on the environmental magnetic interference strength and the structural rules of magnetic interference, the logic of the stage magnetic interference is mined to generate the logic features of environmental magnetic interference. The magnetic field monitoring curve is adaptively filtered and optimized according to the logic features of environmental magnetic interference, and an adaptively filtered and optimized monitoring curve is obtained.

[0024] In this embodiment, the continuously collected global magnetic field monitoring curve of the permanent magnet servo motor rotor is time-sequentially divided, and a plurality of representative time periods are extracted to facilitate subsequent analysis of the timeliness and periodicity of environmental magnetic interference. The magnetic field monitoring curve is a data set obtained by real-time magnetic flux density sampling at a high frequency (such as 2 kHz) through a three-axis Hall array under the motor operating state. First, the overall time sequence is preprocessed to exclude the intense disturbance waveband during the start-up and shutdown process to ensure the accuracy of the stable working condition analysis. Then, the sliding window method (the typical window length is set to 200 ms, and the overlap rate is 50%) is used for time period division, and each piece of data is assigned an independent time label for locating the specific time of interference occurrence. For the region with significant fluctuations, the system will trigger a high-resolution sampling mechanism to shorten the window length to 100 ms to improve the abnormality recognition sensitivity. The multi-time period extraction strategy ensures the full coverage of the magnetic field data under different operating states (steady state, transition, transient state), laying a solid foundation for subsequent interference modeling. The magnetic field monitoring curves of the plurality of time periods extracted above are subjected to interference anomaly recognition to determine whether they are affected by external magnetic fields or internal electromagnetic noise. The recognition method is based on the combination of time-frequency joint analysis and abnormality detection algorithm, and the specific process is as follows: first, Hilbert-Huang transform (HHT) is performed to extract intrinsic mode functions (IMF) and identify high-frequency disturbance characteristics; then, the improved Z-score algorithm is used to statistically evaluate the amplitude change of the high-frequency component of each time period, and if it exceeds the dynamic threshold (generally set to mean ± 3σ), it is determined as an interference fluctuation point. At the same time, a spatial auxiliary judgment mechanism is introduced to further eliminate non-interference normal fluctuations by comparing the consistency of the magnetic flux changes of multiple sensors at the same time. The experiment shows that the recognition sensitivity based on the HHT+Z-score combined method is higher than 90%, effectively marking the time points of power coupling, cable induction, adjacent equipment interference, and other magnetic interference fluctuations, and inputting them as time stamps into the subsequent modeling module.

[0025] Based on the identified magnetic interference fluctuation time points, the magnetic interference intensity at each time point is calculated, and the interference conditions in different time periods are statistically distributed. The magnetic interference intensity calculation adopts the disturbance amplitude integration method, that is, the deviation amplitude of the magnetic flux density signal in the ±10 ms time window near each time point from the mean value is integrated, and normalized processing is performed, expressed in mT (millitesla) units. By statistical analysis of the peak value, mean value and standard deviation of the interference intensity in each time period, the interference intensity distribution sequence is constructed. Further map it to a two-dimensional time-intensity distribution diagram, namely "multi-period interference distribution diagram", the horizontal axis is time, the vertical axis is disturbance intensity, and the interference level is expressed by color gradient (such as blue-low, red-high). In the experiment, it is observed that the magnetic interference increases in the fixed time period of each day (such as 12:00-13:00 noon), which is related to the electromagnetic leakage of the workshop equipment. The distribution diagram can directly reflect the timing law of the interference behavior, and provide visual support for subsequent rule identification. Through data mining and pattern recognition technology, the structured timing law of magnetic interference is extracted from the interference distribution diagram. Self-organizing mapping neural network (SOM) is used for cluster analysis of multi-period interference data, and several typical interference patterns (such as short-time burst type, periodic oscillation type, long-time continuous type, etc.) are identified. Then, combined with the time stamp data, periodic analysis (using fast Fourier transform FFT) is performed to extract the interference frequency component and judge whether there is coupling interference related to the external power frequency (such as 50 Hz). Further, the hidden Markov model (HMM) is used to model the timing transition law of the interference, and the transition probability between interference states and the potential hidden state are identified. In the experiment, it is identified that some interference has a regularity on weekdays (the interference intensity is higher from Monday to Friday, and significantly decreases on weekends), which indicates that the magnetic environment in the industrial scene is closely related to the production rhythm. These structured laws provide a basis for predicting future interference trends and interference source positioning.

[0026] The interference intensity and structural rules are further refined into a set of logical features that can be used for magnetic field data processing. The specific method is to model the multi-dimensional parameters such as interference intensity, frequency component, and time distribution through logistic regression and decision tree model, to generate stage interference trigger conditions and response patterns. For example, the system automatically identifies the "high frequency + high intensity + noon time period" combination as a sudden interference feature; while "more than 10 minutes + stable high amplitude" is marked as a power frequency source coupling feature. These logical features are embedded in the subsequent filtering module in the form of symbolic expressions or conditional rules, to achieve classification and processing priority sorting of different types of interference. Experimental data show that the processing flow assisted by the introduction of logical features improves the accuracy of anomaly identification by about 12% compared with the pure filtering algorithm, and is particularly effective in weak interference background. Based on the constructed logical feature set of magnetic interference, the system implements adaptive filtering processing of the original magnetic field monitoring curve driven by logic. The filtering algorithm selection varies according to the type of interference: for periodic interference, a notch filter is used for targeted suppression; for transient interference, an adaptive Kalman filter is used, which dynamically adjusts the covariance matrix combined with real-time signal statistical characteristics; for complex multi-source interference, a hybrid filter combination strategy is used. The logic feature driven system automatically selects the appropriate filter parameters and dynamically switches the filtering strategy, improving the flexibility and robustness of the processing. The optimized magnetic field curve effectively eliminates more than 80% of the environmental interference components while preserving the key features of the original signal such as waveform trend and peak time. The final output of the adaptive filtering optimization curve provides a clean and reliable data input source for subsequent three-dimensional magnetic field modeling and motor parameter optimization analysis.

[0027] In this embodiment, referring to Figure 3 For the detailed implementation step flowchart of step S2, in this embodiment, the detailed implementation steps of step S2 include: Perform multi-scale magnetic circuit slicing on the global magnetic field accurate compensation model to obtain a magnetic field cross section sequence with different radii and angles; Perform regional magnetic circuit structure analysis on the magnetic field cross section sequence to obtain a plurality of regional magnetic flux paths; Perform topological distribution evolution on the plurality of regional magnetic flux paths to construct a magnetic circuit topological network; According to the magnetic circuit topological network, identify the magnetic pole boundary and the magnetic field line path, and perform regional magnetic field logical depth analysis to obtain the regional magnetic field distribution rule; Calculate the geometric shape center of the servo motor rotor and mark the geometric center point; Based on the geometric center point, construct a real-time magnetic field coordinate system; according to the regional magnetic field distribution rule, perform dynamic magnetic field logical mapping modeling on the real-time magnetic field coordinate system to construct a real-time regional magnetic field logical atlas.

[0028] In this embodiment, the constructed global magnetic field accurate compensation model is spatially sliced to analyze the local details of the rotor magnetic field from different scales and angle dimensions. First, according to the structural parameters of the servo motor rotor (such as stator radius R, pole pair number P, and magnetic pole interval θ), a spatial slicing scheme in polar coordinates is established. The slicing strategy takes the geometric center of the rotor as the origin, and sets multiple slicing planes in the radial and angular directions, for example, every 1 mm in the radial direction (from the axis to the air gap outside), and every 5° in the angular direction, finally forming a multi-dimensional magnetic field cross-section matrix. Extract the local magnetic flux density vector field on each cross-section, including the direction, strength, and gradient information of the magnetic flux. This process relies on the magnetic field finite element simulation model for slicing, which can be automatically completed in simulation environments such as COMSOL Multiphysics or ANSYS Maxwell. By observing the cross-section sequence layer by layer, the magnetic saturation area, magnetic flux bending path, and magnetic transition characteristics between different pole regions can be identified. In the experiment, 72 angles x 10 radius layers of magnetic field slicing were completed on a 12-pole rotor prototype, forming a detailed three-dimensional local magnetic field database. Based on the magnetic field cross-section sequence generated in the previous step, the magnetic flux direction and structure in each region are quantitatively analyzed, and multiple clear magnetic flux conduction paths are extracted. First, extract the equal magnetic flux lines for each cross-section, and use the boundary tracking algorithm (such as Marching Squares or Vector Streamline) to track the starting point, directionality, and connectivity of the magnetic flux line path. Combined with the magnetic flux density vector information and material permeability data, identify the main channel and bypass path of the magnetic flux, and form a complete set of magnetic path paths. To improve the accuracy of path recognition, the system introduces a magnetic energy function gradient constraint strategy to ensure that the extracted path truly reflects the energy conduction direction. In the experiment, obvious multi-path distribution characteristics are identified in the air gap region, pole shoe edge, and stator-rotor transition region, especially in the magnetic pole switching position, there is a magnetic flux bifurcation phenomenon, and part of the magnetic flux path overlaps, indicating that there is local magnetic saturation or magnetic leakage phenomenon. The final regional magnetic flux path data lays the foundation for constructing the topological network.

[0029] The extracted magnetic flux paths are mapped to a topological network model with physical attributes. The magnetic circuit topological network is composed of several "node-path-loop", where the node represents the magnetic flux source point, sink point or path intersection, the path is the magnetic flux conduction channel, and the loop represents the complete magnetic flux closed path. The path weight is introduced when building the network, which is calculated by the magnetic flux intensity, path length, cross-sectional area and permeability, etc. parameters, reflecting the size of the path reluctance. The network structure is established by graph theory modeling tools (such as NetworkX), and the evolutionary model is introduced to observe the activation or weakening of the magnetic flux path under different working conditions. Topological evolution analysis shows that when the rotor angle is near the pole pair junction, part of the path node frequently switches the connection relationship, the network connectivity is weakened, and the risk area of magnetic flux asymmetric fluctuation is prompted. The topological network not only provides a means for macro modeling of magnetic flux distribution, but also provides a logical basis for identifying regional magnetic force lines. Based on the established magnetic circuit topological network, the identification of the magnetic pole boundary and the magnetic force line path is carried out, and the distribution rule of the regional magnetic field is deeply mined. By analyzing the path intensive area and path switching point in the topological graph, the position of the magnetic pole boundary can be determined, especially in the magnetic field gradient mutation area, these boundaries are particularly obvious. Then the magnetic force line tracking algorithm (based on Runge-Kutta integral method) is used to extend from the magnetic pole center outward, and the typical magnetic force line bundle flow diagram is reconstructed, revealing the characteristics of magnetic flux concentration and diffusion. Then the rotor space is divided into multiple magnetic characteristic sub-regions (such as pole shoe area, air gap area, pole edge area) through regional division, and the magnetic flux density statistics, gradient change rate calculation, and magnetic resistance path shortest path identification are analyzed for each sub-region. For example, in the air gap edge area, the magnetic flux density is found to change with angle in a sinusoidal waveform, and the gradient peak value appears in the pole edge ± 2° range. Based on these data, a regional magnetic field distribution rule model is constructed to predict the magnetic flux response characteristics under different structural designs or operating conditions.

[0030] In order to establish a unified spatial reference system, firstly, the geometric form of the servo motor rotor is reconstructed in three dimensions, and the geometric center point is determined as the coordinate reference. The reconstruction method adopts three-dimensional CAD model analysis or obtains the actual physical size data through laser scanning measurement to establish the axisymmetric geometric model of the rotor. Through the axis positioning algorithm of the main shaft direction, combined with the uniform angle between the poles, the rotationally symmetric center of the rotor in the X-Y plane is solved, and the Z-axis direction is projected to the central axis plane, so that the geometric center point (marked as O point) is located in three dimensions. The center point as the origin of the magnetic flux direction and path calculation ensures that the calculation results of the magnetic field in different regions are comparable and integrable under the consistency of the coordinates. In the experiment, the error of the marked geometric center is controlled within 0.1 mm, which meets the demand of high-precision magnetic circuit analysis. The magnetic field analysis is combined with the spatial coordinate system to construct a unified and dynamic variable magnetic field logical atlas. On the basis of the geometric center O point, the polar coordinate system (r, θ, z) is constructed, and the magnetic flux vector data is mapped into the coordinate system in real time. Combined with the aforementioned regional magnetic field distribution rule model, the real-time sampling magnetic flux density data is compared and analyzed with the known rule model to realize logical mapping modeling. The model uses the magnetic flux direction-intensity logical contrast matrix to calculate the error between the magnetic flux distribution in each sub-region and the standard distribution model, and real-time displays the regional magnetic flux state deviation. For example, at a specific rotation angle θ=75° and a radius r=25mm position, when the magnetic flux density is 20% lower than the standard model, the system triggers an abnormal alarm, and highlights the region in red in the logical atlas. The atlas supports real-time dynamic updating and has a rotation tracking function, which adapts to the magnetic field changes in the motor operation. Finally, the real-time regional magnetic field logical atlas not only improves the intelligent level of magnetic field monitoring, but also provides data visualization support for motor magnetic circuit design optimization.

[0031] In this embodiment, reference Figure 4 The detailed implementation steps of step S3 include: Deep magnetic field element identification is performed on the global magnetic field accurate compensation model to mark the magnetic poles and magnetic lines to obtain regional magnetic field elements; Dynamic magnetic flux tracking is performed on the regional magnetic field elements to extract the evolution trajectories of multiple magnetic field nodes; Magnetic field state change analysis is performed on the evolution trajectories of multiple magnetic field nodes to obtain the magnetic field trajectory state change trend; According to the magnetic field trajectory state change trend, multi-node change trajectory correlation analysis is performed to obtain the magnetic coupling influence characteristics between different magnetic field nodes; According to the magnetic coupling influence characteristics, periodic magnetic field linkage evolution is performed to construct a multi-node magnetic field coupling dependent network.

[0032] In this embodiment, the established global magnetic field precise compensation model is used to identify the magnetic field elements at the level of deep recognition, so as to accurately extract each key magnetic field composition unit in the rotor structure, including magnetic poles, magnetic lines, magnetic flux distribution belts and the like, and form a complete regional magnetic field element set. The realization path includes two modules of vectorization processing of magnetic field tensor field and magnetic field boundary constraint identification. First, the principal component analysis (PCA) is performed on the three-dimensional magnetic flux density distribution data, and the main magnetic flux direction in each space unit is extracted to construct an initial magnetic line path set. Then, the density segmentation threshold method (the critical density threshold is set to 0.25 T) is used to identify the magnetic pole center area, and the boundary of the magnetic flux density mutation area between regions is depicted to complete the geometric extraction of the magnetic pole boundary. Further, combined with the magnetic flux direction information, the vector integral method based on Runge-Kutta algorithm is used to track the complete magnetic line path from the magnetic pole center point, and the trend and length distribution are recorded. Finally, the regional magnetic field element atlas is formed, including the magnetic pole number, polarity information, force line trend, magnetic pole coverage range and the like, which provides a spatial basis and boundary condition for the magnetic flux evolution trajectory analysis. Based on the magnetic field element set extracted in the last stage, the magnetic field nodes with key magnetic flux flow significance are dynamically tracked, and their spatial evolution trajectories in the rotor running process are recorded. First, the rotating motion boundary condition (the rotating speed is set to 3000 rpm, and the simulation time window is 100 ms) is applied in the simulation environment (such as ANSYS Maxwell or COMSOL) to simulate the running state of the permanent magnet servo motor under normal load. The system takes time sequence as the main axis, continuously extracts the three-dimensional position coordinates, magnetic flux density vector values, gradient change rates and the like of the magnetic field nodes, and forms a time-space-intensity three-dimensional trajectory data set. The nodes to be tracked include: magnetic pole center point, magnetic flux maximum point, inter-pole neutral point and magnetic flux sudden change area and the like. The Lagrangian magnetic flux tracking method is used to ensure the continuity and spatial accuracy of the nodes in the motion process. Through trajectory mapping, it is observed that the magnetic pole edge nodes present a "eccentric running" trajectory, and the magnetic flux dominant points form a regular periodic fluctuation path, which lays a foundation for the magnetic flux coupling analysis.

[0033] The magnetic field node evolution trajectory obtained in the previous step is analyzed to extract its state change characteristics over time and space. First, the system extracts statistical features from each trajectory data sequence, including the mean, peak, rate of change, and periodicity of the magnetic flux density. Second, time-frequency analysis methods such as short-time Fourier transform (STFT) and wavelet packet decomposition (WPD) are introduced to identify the multi-frequency disturbance patterns and dominant frequency responses contained in the trajectory. During the analysis, by comparing the state sequences of multiple nodes, it is found that the magnetic flux at the center of the magnetic pole changes more smoothly, while the magnetic flux density at the pole edge area shows greater volatility, especially near the position of each magnetic pole switching, with a change amplitude of more than 0.12 T. At the same time, the node magnetic flux intensity waveform shows a clear sinusoidal modulation trend, consistent with the magnetic coupling characteristics in motor operation. Finally, the state change trend map of each magnetic field node is constructed, and indexes such as "stability factor" and "disturbance coefficient" are defined to quantify the stability of the magnetic field behavior. Around the state change trend of the magnetic field trajectory, cross-comparison analysis between multiple nodes is performed to identify the potential magnetic coupling influence relationship. The system uses multivariate time series cross-correlation function (CCF) and Granger causality analysis methods to model the time lag correlation between the change trends of each magnetic field node, and extracts the magnetic coupling strength and coupling directionality. For example, when the magnetic flux peak fluctuation occurs at the edge of the magnetic pole node, the adjacent magnetic pole center node responds with a magnetic flux rise within 10 ms, indicating a strong correlation. Further, a magnetic flux path disturbance transmission model is introduced to simulate the flow and propagation process of magnetic field disturbances in magnetic lines. Experimental data show that there is a clear signal transmission path between nodes with a magnetic flux change rate greater than 3 T / s, forming a magnetic flux coupling chain. Finally, the system outputs a set of magnetic coupling characteristic matrices, including coupling strength coefficient (value range 0~1), response time delay, spatial distance, etc., revealing the dynamic coupling relationship between different nodes inside the magnetic field, and providing a theoretical basis for resonance avoidance and disturbance regulation in electromagnetic design.

[0034] A periodic linkage evolution model between multiple nodes, i.e., a multi-node magnetic field coupling dependency network, is constructed in combination with the magnetic coupling feature matrix. The network takes the magnetic field nodes as the "nodes" in the graph structure and the coupling relationship as the "edge". The edge weight is calculated by the coupling strength coefficient and the response time delay. The weighted directed graph modeling method is adopted to express the directionality and hierarchy of the magnetic flux influence. In the network evolution simulation process, the periodic boundary condition is introduced to map the working period of the servo motor rotor (such as 20 ms period corresponding to 3000 rpm) to the magnetic flux propagation period, and analyze the stability and variability of the coupling relationship in the continuous period. Network analysis shows that there is a highly stable periodic linkage relationship between the magnetic pole center node and the pole edge node, while the coupling imbalance phenomenon caused by magnetic field shift occurs between the asymmetrically arranged pole pairs. In addition, the PageRank algorithm is used to identify the key nodes that have the greatest impact on the global magnetic flux, providing input basis for subsequent magnetic field compensation strategy design. The finally formed magnetic field coupling dependency network atlas not only improves the understanding of the complex magnetic flux conduction mechanism, but also supports the subsequent electromagnetic optimization design and motor operation anomaly prediction work.

[0035] In this embodiment, step S4 includes the following steps: Trajectory trend spatial distribution analysis is performed on the magnetic field trajectory state change trend to obtain a multi-node magnetic field trend distribution map; Based on the multi-node magnetic field coupling dependency network, a short-term magnetic field evolution prediction is performed on the multi-node magnetic field trend distribution map to generate a short-term magnetic field prediction distribution map; According to the short-term magnetic field prediction distribution map, a regional magnetic flux density value calculation is performed to obtain a short-term magnetic field evolution density prediction value; Magnetic intensity time sequence distribution fluctuation mining is performed on the short-term magnetic field prediction distribution map to generate a magnetic intensity time sequence distribution fluctuation feature; According to the short-term magnetic field evolution density prediction value and the magnetic intensity time sequence distribution fluctuation feature, a comprehensive electromagnetic performance evaluation is performed to obtain an electromagnetic performance evaluation report.

[0036] In this embodiment, by reconstructing and clustering the spatial dimension of the trend of the previously extracted magnetic field trajectory state change, the specific distribution pattern of the magnetic flux evolution trend in space is identified. In specific implementation, the system first normalizes the magnetic flux change trend of all key magnetic field nodes (including the magnetic pole center, the magnetic pole edge, the stator tooth top, the tooth neutral zone, etc.) in the running period, such as normalizing the amplitude range to [0, 1] and removing high-frequency noise interference (>1 kHz component). Then, using the joint modeling method based on principal component analysis (PCA) and K-means clustering algorithm, the magnetic flux time series trend vector of each node is reduced in dimension and clustered in space, and finally multiple change trends are classified into representative patterns (such as linear enhancement type, periodic swing type, asymmetric jump type, etc.). In the spatial mapping stage, a three-dimensional coordinate framework is constructed using the rotor geometric model, and the trend pattern of each node is mapped back to the original physical space. The finally generated “multi-node magnetic field trend distribution map” presents the distribution characteristics of the magnetic flux trend type in each region in a chromatographic manner, which helps to identify the local abnormal area, coupled cluster and fluctuation integrated source and other structural characteristics of the magnetic field behavior from the spatial dimension. After completing the construction of the magnetic field trend spatial distribution map, the next task is to use the established multi-node magnetic field coupling dependence network to predict the short-term magnetic field trend, and to realize the advance judgment of the magnetic flux behavior in the future one running period (such as 20ms, based on 3000rpm running state). First, the system constructs a short-term magnetic flux trend prediction model based on the recurrent neural network (RNN) structure, especially the time-gated LSTM (Long Short-Term Memory) unit. The model input is the magnetic flux state change data of each node at the current time and the historical n time slices (n=20, corresponding to 1ms resolution in 20ms), combined with the synchronous state data of the topological neighbor nodes in the magnetic coupling network. In the training stage, historical simulation data or actual sampling data is used for supervised learning of the prediction model, and the mean square error (MSE) index is selected as the error function, with the goal of controlling the prediction error to be less than 0.01T. After training, the prediction module is run in the real-time simulation or virtual simulation system, and the predicted value of the magnetic flux density evolution of each node in the future period is output. The prediction result is mapped in three-dimensional space to form a “short-term magnetic field prediction distribution map”, which serves as a spatial basis for subsequent density prediction and performance evaluation.

[0037] On the basis of the magnetic field trend prediction, the quantitative magnetic flux density numerical calculation is carried out for each spatial region of the servo motor rotor to obtain the quantitative indexes of the magnetic flux level of each local region in the prediction time window. The system first divides the entire rotor working domain into multiple structural functional partitions, such as the pole face center area, the pole edge transition area, the air gap middle part, the tooth top concentrated area and the like. Each partition is delimited by a three-dimensional space grid, and the spatial scale accuracy is controlled to be 0.5 mm cubic unit. According to the magnetic flux values of the grid nodes in the “short-term magnetic field prediction distribution map”, the system calculates the average value, the maximum value, the minimum value and the spatial gradient amplitude of the magnetic flux density in each region, and outputs the “short-term magnetic field evolution density prediction value matrix”. Under the experimental conditions (such as the simulation temperature is set to 75°C, and the material attribute is NdFeB magnetic steel B=1.2T), the simulation results show that the density fluctuation of the pole face edge region is larger, and the density gradient amplitude reaches 0.25T / mm, while the change amplitude of the pole center region is smaller, only 0.03T / mm. These predicted density values will be used as the key input variables for judging the magnetic saturation, the magnetic hysteresis and the power output capability.

[0038] After obtaining the sub-region density prediction values, the system further analyzes the fluctuation behavior of the magnetic intensity over time, excavates its time series distribution fluctuation law, and reveals the potential structural disturbance mechanism. The methods used include Fourier transform (FFT) to identify the main frequency components, wavelet decomposition (such as Daubechies 4th order) to capture instantaneous changes, and moving average filtering to extract the stationary trend. After processing by these methods, the system generates a "magnetic intensity fluctuation spectrum" for each region, which contains key parameters such as main frequency distribution, periodic fluctuation index (PWI), and peak / valley time points. In the analysis process, it is found that for the edge regions of the magnetic pole, the 50Hz and 150Hz harmonic components are significantly enhanced, with fluctuation amplitudes reaching more than 20% of the original signal; while in the central region of the magnetic pole, the spectrum distribution is concentrated in the low frequency band, with less fluctuation. In addition, the system also calculates the "magnetic flux stability coefficient" (MFS) of each region, defined as the inverse of the variance of the magnetic flux density change per unit time, for quantifying the stability of the magnetic field. The final "magnetic intensity time series fluctuation characteristic table" is an important reference for the next step of comprehensive performance evaluation. Based on all the intermediate analysis results above, the system systematically evaluates the current and short-term electromagnetic performance of the rotor. The system takes the "short-term magnetic field evolution density prediction value matrix" and the "magnetic intensity time series fluctuation characteristic table" as input data, constructs a set of multi-index fusion models, and evaluates the dimensions including: magnetic flux utilization ratio (Magnetic Utilization Ratio, MUR), magnetic saturation probability index (MSPI), magnetic flux stability index (Magnetic Field Stability Index, MFSI), and magnetic field uniformity coefficient (Magnetic Uniformity Coefficient, MUC). Each index is graded by setting a threshold, for example, MUR higher than 90% is considered excellent, and MSPI lower than 5% indicates low saturation risk. Finally, the system conducts a comprehensive scoring of all indexes based on a weighted decision model, and outputs the structural performance grade (such as A+, A, B, etc.). In addition, the report also includes magnetic field anomaly warning areas, suggestions for magnetic pole shape optimization (such as moderately increasing the pole face chamfer radius), and dynamic compensation suggestions. The comprehensive electromagnetic performance evaluation report provides strong data support and decision basis for rotor structure design optimization, operation state adjustment, and predictive maintenance.

[0039] In this embodiment, step S5 includes the following steps: Real-time magnetic field transient mutation recognition is performed according to the global magnetic field accurate compensation model, and the magnetic field transient mutation point is marked; The magnetic field transient mutation point is subjected to transient mutation tracing and positioning, and the local magnetic field transient mutation region is extracted; According to the real-time regional magnetic field logical atlas, the magnetic field transient mutation region is subjected to magnetic field mutation diffusion evolution, and the magnetic field mutation diffusion evolution characteristics are obtained; The diffusion distribution model of the magnetic field mutation is obtained by tracking the space-time evolution of the diffusion evolution characteristics of the magnetic field mutation and visualizing rendering.

[0040] In this embodiment, by analyzing the real-time output magnetic field data in the global magnetic field precise compensation model, the sudden change of magnetic flux in the working process of the rotor is identified, and the spatial position is marked with high precision. First, the system samples the real-time magnetic flux data of the rotor in the running state at a high frequency, and the sampling frequency is controlled above 20 kHz to ensure the response capture accuracy of the short-time transient magnetic field behavior. In the data processing stage, an improved sliding window mutation detection algorithm (such as CUSUM enhanced mutation detection) is used, with every 10 ms as the analysis window, to calculate the first and second derivatives of the magnetic flux density change per unit time. If the change rate exceeds the set threshold (for example, the magnetic flux change exceeds 0.15 T / ms or the derivative change rate > 300%), the system determines that it is a mutation event, and records the point as a "magnetic transient mutation point". In the actual evaluation process, the common mutation points are mostly concentrated in the magnetic pole switching area, the air gap end or the magnetic pole transition zone, and the spatial markers are stored in three-dimensional polar coordinates, including polar radius r, angle θ and axial z distribution, so as to facilitate subsequent spatial positioning and diffusion analysis. Further trace the source of the mutation event and its influence range, and construct the spatial description of the mutation area. For this purpose, the system uses a magnetic field mutation source model based on causal convolution (Causal Convolution) network to deeply analyze the magnetic flux time series of the time period before and after the mutation point. The specific method is to extract the 20 ms magnetic field change sequence before and after the mutation point, and analyze whether the change pattern is associated with structural defects, magnetic interference or current driving instability through model analysis. In the spatial dimension, the local magnetic field gradient inversion technology is used to calculate the gradient change rate of the magnetic flux distribution in the surrounding area of the mutation point to identify the diffusion path of the mutation source. For example, experimental data shows that within 1 mm radius around the mutation point, if the magnetic field gradient reaches 0.5 T / mm and continuously strengthens in a certain direction, it can be inferred that this direction is the main diffusion direction. Finally, the system takes the mutation point as the center, combines the magnetic flux gradient contour to construct a "local magnetic transient mutation area", and outputs it in the form of a three-dimensional grid as the basic input of the evolution modeling.

[0041] The magnetic flux diffusion path in the mutation region is dynamically modeled and the evolution characteristics are analyzed using the previously constructed "real-time magnetic field logical map". The real-time magnetic field logical map contains the magnetic flux direction, coupling strength, magnetic pole boundary response characteristics, etc. of each spatial node. The magnetic flux diffusion direction is simulated by a field line tracing algorithm. The system takes the mutation point as the initial seed point and simulates the magnetic flux evolution in multiple directions around it. The diffusion rate is dynamically adjusted according to the magnetic resistance and material permeability (such as NdFeB region μr=1.05, iron core region μr=5000) in the diffusion path. During the diffusion evolution process, the system extracts multiple physical quantities such as mutation duration, diffusion area, average magnetic flux density in the diffusion path, and fluctuation frequency to construct the "magnetic field mutation diffusion evolution feature set". For example, the typical mutation diffusion time is between 25ms, the spatial expansion range is usually within 36mm, and the evolution feature set reflects that the air gap magnetic flux density change rate is as high as 0.2T / ms, indicating the risk of mutation caused by electromagnetic oscillation or structural resonance. The above magnetic flux diffusion evolution process is reconstructed in the time and space dimensions, and an interactive visual distribution model is constructed to serve engineering diagnosis and magnetic circuit design optimization. The system uses four-dimensional visual modeling technology to display the magnetic flux diffusion process in three-dimensional space (r, θ, z) and time t dimension. The mutation propagation path is dynamically displayed using volume rendering (Volume Rendering) and time axis sliding. The key visualization dimensions include: magnetic flux intensity color mapping, diffusion speed arrow vector, mutation source position flickering marker, etc. The rendering tool supports WebGL or OpenGL platform, compatible with CAD environment and magnetic field simulation platform (such as ANSYS Maxwell or JMAG). In actual engineering, this model supports operators to select, measure and simulate the mutation affected area, such as analyzing whether the mutation will cause local overheating of the winding, magnetic vibration noise rise and other side effects. The final "magnetic field mutation diffusion distribution model" not only directly reveals the dynamic process of the mutation event, but also provides decision support for further optimization of magnetic circuit design, control strategy adjustment and fault prediction.

[0042] In this embodiment, step S6 includes the following steps: The magnetic field mutation diffusion distribution model is subjected to mutation diffusion attribution mining, and the mutation diffusion factor is extracted; The diffusion range situation and diffusion time of the mutation diffusion factor are calculated to obtain the multi-dimensional state of the mutation diffusion; Based on the electromagnetic performance evaluation report, the magnetic field mutation control iteration simulation is performed on the multi-dimensional state of the mutation diffusion to generate multiple magnetic field mutation control schemes; According to the multiple magnetic field mutation control schemes, the quantitative calculation of the control effect of each scheme is performed to obtain the control effect value of each scheme; Based on the control effect value, the optimal magnetic field mutation control scheme is extracted; Intelligent parameter optimization planning is performed according to the optimal magnetic field mutation regulation scheme, so as to obtain an optimal magnetic field parameter configuration; The optimal magnetic field parameter configuration is used to perform intelligent regulation optimization of the permanent magnet servo motor.

[0043] In this embodiment, backtracking reason mining is performed based on a mutation diffusion distribution model. The system uses a combination of random forest and causal Bayesian network technology to extract dominant factors from a large number of mutation event samples. For example, a multi-variable sorting is performed in combination with parameters such as magnetic flux gradient, material saturation, and rotational speed deviation to calculate the “importance score” (0-1 interval) of each type of factor. In the experiment, the magnetic pole edge passivation, the increase of the tooth slot effect to the conduction path length, and the 30 μm shaft center eccentricity are high importance influencing factors; these factors are refined as “magnetic flux interference structure defect factors”. This attribution analysis provides a directional marker for mutation control. After refining the key factors, the influence of the key factors on the diffusion range and speed of the mutation is quantitatively measured. The system iteratively tests each mutation event through numerical simulation combined with a statistical regression model, and monitors the diffusion space and time by changing a single factor (such as a 10% decrease in magnetic permeability and a ±5% motor speed). A multi-dimensional state table is established: for a certain structure defect factor, the downstream diffusion range increases by 30%, and the diffusion time is shortened to 2 ms; while a 10°C increase in ring temperature expands the diffusion range by 10% and delays the time by 1 ms. The “mutation diffusion multi-dimensional state library” finally output can be directly used for simulating the mutation behavior exit mechanism or designing the regulation requirements.

[0044] Relying on the structure performance report and the multi-dimensional state library, the system establishes a regulation simulation platform to simulate the control parameters of each mutation factor. Parameter combinations (such as compensation pole thickness ±0.05 mm, pole arc shape correction ±2°, and stator interference angle adjustment ±0.5°) are selected by Monte Carlo sampling. Through hundreds of simulation iterations, multiple regulation schemes are generated, including “pole chamfer fine tuning”, “local magnetic field enhancement superposition”, and “air gap correction”. Each scheme calculates the mutation diffusion change indicators in a three-dimensional magnetic field simulation environment, such as diffusion zone radius, duration, harmful gradient peak value, and structure stability change, laying a foundation for quantitative evaluation in the next step. The effect of each regulation scheme is objectively quantified, and the control results are evaluated through a series of indicators. Key parameters include mutation frequency reduction rate, average diffusion radius contraction ratio, structure integrity improvement degree, and operating efficiency change. The system compares the simulation model magnetic flux distribution, gradient atlas, and instability timing characteristics before and after the scheme. For example, under the “pole chamfer + interference angle adjustment”, the mutation frequency is reduced by 40%, the diffusion radius is reduced by 25%, the structure integrity is improved by 7%, and the system efficiency is improved by 0.5%. These indicators are combined to form a single “regulation effect value”, and a horizontal comparison matrix of multiple schemes is provided.

[0045] After obtaining the reference scheme, the system generates an optimized parameter space and performs intelligent optimization planning. A constraint optimization algorithm (such as genetic algorithm GA or particle swarm optimization PSO) is used to iteratively fine-tune key parameters (pole chamfer radian, pole arc ratio, magnetic material thickness, air gap average deviation). The optimization goal is a dual objective function of minimizing extreme mutation and maximizing efficiency. The system can suggest: "increase the pole chamfer from 0.5mm to 0.53mm, adjust the pole arc ratio by +1.2°, and adjust the air gap average by -0.02mm" combination to minimize the mutation diffusion time + expansion. In another scheme, it is suggested to add "local steel sheet reinforcement" and other structural reinforcement methods. Finally, the optimal parameter configuration is issued to the design or manufacturing process, such as modifying the pole shape geometry in CAD / CAM or adjusting the pole spacing and pole shape in assembly to ensure air gap consistency and structural stability. After manufacturing, the actual test verifies the improvement of magnetic flux mutation behavior, and the controller software uploads the parameters and monitors the real-time effect to ensure the stability of the optimization results. The system can continuously monitor for more than 100 hours, and under the conditions of hot start or load change, the mutation diffusion phenomenon continues to decrease by 30%+, the efficiency increases by 0.3-0.7%, the structural reliability increases by 20%, and the life expectancy is extended by 5%.

[0046] In the present embodiment, a magnetic field simulation and parameter optimization system for a permanent magnet servo motor rotor is provided for performing the magnetic field simulation and parameter optimization method for a permanent magnet servo motor rotor as described above, comprising: A magnetic field distribution calculation module is configured to collect the rotor global magnetic field monitoring curve during the operation of the servo motor rotor, and perform real-time magnetic flux density gradient calculation and local magnetic field compensation modeling to construct a global magnetic field accurate compensation model. A magnetic field logic analysis module is configured to analyze the regional magnetic circuit structure of the global magnetic field accurate compensation model, and perform dynamic magnetic field logic mapping modeling to construct a real-time regional magnetic field logic map. A trajectory correlation module is configured to analyze the magnetic field state change of the global magnetic field accurate compensation model, and perform multi-node change trajectory correlation analysis to construct a multi-node magnetic field coupling dependency network. A magnetic field evolution prediction module is configured to perform short-term magnetic field evolution prediction based on the multi-node magnetic field coupling dependency network, and perform comprehensive electromagnetic performance evaluation to obtain an electromagnetic performance evaluation report. A mutation diffusion distribution module is configured to identify real-time magnetic field transient mutation based on the real-time regional magnetic field logic map, and perform time-space evolution tracking to obtain a magnetic field mutation diffusion distribution model. A magnetic field parameter tuning module is configured to perform magnetic field mutation regulation and control iterative simulation based on the electromagnetic performance evaluation report and the magnetic field mutation diffusion distribution model, and perform intelligent parameter tuning planning to execute intelligent regulation and control optimization of the permanent magnet servo motor.

[0047] Therefore, the embodiments should be regarded, at any point, as being exemplary and not limiting, the scope of the application being defined by the appended claims and not by the above description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

[0048] The foregoing merely illustrates the principles of the application and applies only to the particular cases described and illustrated herein. Numerous modifications in the embodiments described herein will be readily apparent to those skilled in the art, and the principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Accordingly, the application is not to be restricted as illustrated in the foregoing description, but is to be accorded the full scope consistent with the claims made to the principles and novel features disclosed herein.

Claims

1. A magnetic field simulation and parameter optimization method for a permanent magnet servo motor rotor, characterized in that: The following steps are involved: Step S1: Collect the global magnetic field monitoring curve of the servo motor rotor during operation, perform real-time magnetic flux density gradient calculation and local magnetic field compensation modeling, and build a global magnetic field precision compensation model; Step S2: Perform regional magnetic circuit structure analysis on the global magnetic field precision compensation model, perform dynamic magnetic field logic mapping modeling, and construct a real-time regional magnetic field logic map; Step S3: Analyze the magnetic field state change of the global magnetic field precision compensation model, and perform multi-node change trajectory correlation analysis to construct a multi-node magnetic field coupling dependency network; Step S4: performing short-term magnetic field evolution prediction based on the multi-node magnetic field coupling dependency network and performing comprehensive electromagnetic performance evaluation to obtain an electromagnetic performance evaluation report; Step S5: identifying real-time magnetic field transient mutations based on the real-time regional magnetic field logic map, and tracking the spatiotemporal evolution to obtain a magnetic field mutation diffusion distribution model; Step S6: Perform an iterative simulation of magnetic field mutation control based on the electromagnetic performance evaluation report and the magnetic field mutation diffusion distribution model, and perform intelligent parameter tuning planning to execute the intelligent control optimization operation of the permanent magnet servo motor.

2. The magnetic field simulation and parameter optimization method for a permanent magnet servo motor rotor according to claim 1, characterized in that: The specific steps of step S1 are: Detect the operating state of the servo motor rotor, perform three-dimensional magnetic field scanning, and obtain the rotor's full-area magnetic field monitoring curve; Perform adaptive magnetic interference filtering optimization on the rotor global magnetic field monitoring curve to obtain an adaptive filtering optimization monitoring curve; Perform three-dimensional spatial distribution analysis on the adaptive filtering optimization monitoring curve to obtain the rotor's full-domain magnetic field distribution map; Perform real-time magnetic flux density gradient calculation based on the rotor's full-area magnetic field distribution map to generate a magnetic flux density gradient map; Perform spatial frequency domain transformation on the magnetic flux density gradient map to construct the magnetic field frequency domain characteristic spectrum; The magnetic field frequency domain characteristic spectrum is decomposed layer by layer, and the interlayer magnetic flux coupling is calculated to obtain the multi-layer magnetic flux coupling parameters; The interlayer coupling difference analysis is performed based on the multi-layer magnetic flux coupling parameters, and local magnetic field compensation modeling is performed to construct a global magnetic field precision compensation model.

3. The magnetic field simulation and parameter optimization method for a permanent magnet servo motor rotor according to claim 2, characterized in that: The specific steps of performing adaptive magnetic interference filtering optimization on the rotor global magnetic field monitoring curve to obtain the adaptive filtering optimization monitoring curve are: Decompose the rotor global magnetic field monitoring curve into multiple time periods and extract the magnetic field monitoring curves of multiple time periods; Identify environmental magnetic interference based on the magnetic field monitoring curve and mark the time point of magnetic interference fluctuation; The environmental magnetic interference intensity is calculated according to the time point of magnetic interference fluctuation, and the multi-period interference distribution analysis is performed to obtain the multi-period interference distribution map; The structured laws of environmental magnetic interference are identified on the multi-period interference distribution map to obtain the structured laws of magnetic interference; Based on the intensity of environmental magnetic interference and the structured laws of magnetic interference, the logic mining of staged magnetic interference is carried out to generate the logical features of environmental magnetic interference; The magnetic field monitoring curve is adaptively optimized by magnetic interference filtering according to the logical characteristics of the environmental magnetic interference, thereby obtaining an adaptive filtering optimized monitoring curve.

4. The magnetic field simulation and parameter optimization method for a permanent magnet servo motor rotor according to claim 1, characterized in that: The specific steps of step S2 are: Perform multi-scale magnetic circuit slicing on the global magnetic field precision compensation model to obtain a sequence of magnetic field cross sections with different radii and angles; Perform regional magnetic circuit structure analysis on the magnetic field cross-section sequence to obtain multiple regional magnetic flux paths; Perform topological distribution evolution on multiple regional magnetic flux paths to construct a magnetic circuit topology network; Identify magnetic pole boundaries and magnetic field lines based on the magnetic circuit topology network, and perform in-depth analysis of the magnetic field logic within the region to determine the magnetic field distribution pattern within the region; Calculate the geometric center of the servo motor rotor and mark the geometric center point; Construct a real-time magnetic field coordinate system based on the geometric center point; According to the magnetic field distribution law in the region, the real-time magnetic field coordinate system is dynamically mapped and modeled with magnetic field logic to construct a real-time regional magnetic field logic map.

5. The magnetic field simulation and parameter optimization method for a permanent magnet servo motor rotor according to claim 1, characterized in that: The specific steps of step S3 are: Perform deep magnetic field element identification on the global magnetic field precision compensation model, mark magnetic poles and magnetic lines of force, and obtain regional magnetic field elements; Perform dynamic magnetic flux tracking on regional magnetic field elements and extract the evolution trajectory of multiple magnetic field nodes; Analyze the magnetic field state changes of the evolution trajectories of multiple magnetic field nodes to obtain the magnetic field trajectory state change trend; According to the trend of magnetic field trajectory state change, multi-node change trajectory correlation analysis is performed to obtain the magnetic coupling influence characteristics between different magnetic field nodes; According to the magnetic coupling influence characteristics, periodic magnetic field linkage evolution is performed to construct a multi-node magnetic field coupling dependency network.

6. The magnetic field simulation and parameter optimization method for a permanent magnet servo motor rotor according to claim 1, characterized in that: The specific steps of step S4 are: Perform trajectory trend spatial distribution analysis on the magnetic field trajectory state change trend to obtain a multi-node magnetic field trend distribution map; Based on the multi-node magnetic field coupling dependency network, the short-term magnetic field evolution prediction of the multi-node magnetic field trend distribution map is performed to generate a short-term magnetic field prediction distribution map; Calculate the regional magnetic flux density value according to the short-term magnetic field prediction distribution map to obtain the short-term magnetic field evolution density prediction value; Mining the magnetic intensity time series distribution fluctuations on the short-term magnetic field prediction distribution map to generate the magnetic intensity time series distribution fluctuation characteristics; A comprehensive electromagnetic performance evaluation is conducted based on the short-term magnetic field evolution density prediction value and the magnetic intensity time series distribution fluctuation characteristics to obtain an electromagnetic performance evaluation report.

7. The magnetic field simulation and parameter optimization method for a permanent magnet servo motor rotor according to claim 1, characterized in that: The specific steps of step S5 are: Based on the global magnetic field precision compensation model, real-time magnetic field transient mutation identification is performed to mark the magnetic field transient mutation points; Tracing and locating the transient mutation points of the magnetic field and extracting the local magnetic field transient mutation area; According to the real-time regional magnetic field logic map, the magnetic field mutation diffusion evolution is carried out on the local magnetic field transient mutation area to obtain the magnetic field mutation diffusion evolution characteristics; The spatiotemporal evolution of the magnetic field mutation diffusion evolution characteristics is tracked and visualized to obtain the magnetic field mutation diffusion distribution model.

8. The magnetic field simulation and parameter optimization method for a permanent magnet servo motor rotor according to claim 1, characterized in that: The specific steps of step S6 are: Conduct mutation diffusion attribution mining on the magnetic field mutation diffusion distribution model and extract mutation diffusion factors; Calculate the diffusion range and diffusion time of the mutation diffusion factor to obtain the multi-dimensional state of mutation diffusion; Based on the electromagnetic performance evaluation report, the multi-dimensional state of mutation diffusion is iteratively simulated to control the magnetic field mutation, and multiple magnetic field mutation control schemes are generated; According to multiple magnetic field mutation control schemes, quantitative calculation of the control effect of each scheme is performed to obtain the control effect value of each scheme; Extracting an optimal magnetic field mutation control scheme based on the control effect value; Perform intelligent parameter tuning planning based on the optimal magnetic field mutation control scheme to obtain the optimal magnetic field parameter configuration; Configure the optimal magnetic field parameters to perform intelligent control optimization of permanent magnet servo motors.

9. A magnetic field simulation and parameter optimization system for a permanent magnet servo motor rotor, characterized in that: The method for performing magnetic field simulation and parameter optimization for a permanent magnet servo motor rotor according to claim 1 comprises: The magnetic field distribution calculation module is used to collect the global magnetic field monitoring curve of the servo motor rotor during operation, and perform real-time magnetic flux density gradient calculation and local magnetic field compensation modeling to build a global magnetic field precision compensation model; The magnetic field logic analysis module is used to analyze the regional magnetic circuit structure of the global magnetic field precision compensation model, perform dynamic magnetic field logic mapping modeling, and construct a real-time regional magnetic field logic map; The trajectory correlation module is used to analyze the magnetic field state changes of the global magnetic field precision compensation model, perform multi-node change trajectory correlation analysis, and build a multi-node magnetic field coupling dependency network; The magnetic field evolution prediction module is used to predict the short-term magnetic field evolution based on a multi-node magnetic field coupling dependency network, and conduct a comprehensive electromagnetic performance evaluation to obtain an electromagnetic performance evaluation report; The mutation diffusion distribution module is used to identify real-time magnetic field transient mutations based on the real-time regional magnetic field logic map, and track the spatiotemporal evolution to obtain a magnetic field mutation diffusion distribution model; The magnetic field parameter tuning module is used to perform iterative simulation of magnetic field mutation control based on the electromagnetic performance evaluation report and the magnetic field mutation diffusion distribution model, and to carry out intelligent parameter tuning planning to execute the intelligent control optimization operation of the permanent magnet servo motor.

Images