Sectional sinusoidal permanent magnet optimization method of permanent magnet synchronous motor

Abstract

The invention discloses a sectional sinusoidal permanent magnet optimization method for a permanent magnet synchronous motor. Finite element software is used for establishing a standard surface-mounted permanent magnet synchronous motor model; applying a mathematical method of linear interpolation to a sine curve, changing the shape of a surface-mounted permanent magnet into the shape of a sectional sinusoidal permanent magnet, replacing the permanent magnet in the surface-mounted permanent magnet synchronous motor model with the sectional sinusoidal permanent magnet, and establishing a sectional sinusoidal permanent magnet synchronous motor model; the position of a point on a sine curve of a sectional sinusoidal permanent magnet is used as an optimization variable, the cogging torque of the permanent magnet synchronous motor and the vibration noise of the motor in a rated operation state are used as optimization targets, and the optimal position of the point on the sine curve is searched by using a method of combining a convolutional neural network proxy model and a multi-target genetic algorithm. And optimization of the sectional sinusoidal permanent magnet is realized. According to the permanent magnet synchronous motor, the cogging torque is reduced, the vibration noise and the use amount of the permanent magnets in the permanent magnet synchronous motor are reduced, and the utilization efficiency of the permanent magnets is improved.

Description

Technical Field

[0001] The invention relates to the design field of permanent magnets in permanent magnet synchronous motors, and in particular to a segmented sinusoidal permanent magnet optimization method for permanent magnet synchronous motors. Background Art

[0002] In recent years, with the vigorous development of power electronics technology and new energy vehicles, permanent magnet synchronous motors, as a common type of motor in the field of new energy vehicles, have become an important topic in new energy vehicles and my country's industrial field.

[0003] Permanent magnet synchronous motor has the advantages of high efficiency, simple structure, large torque and high reliability. However, during the operation of permanent magnet synchronous motor, permanent magnet material will generate cogging torque and vibration noise due to vibration, high temperature and overload current. On the one hand, changing the structure and shape of permanent magnet synchronous motor will affect the magnetic field distribution in the air gap of the motor. Therefore, in the design process of permanent magnet structure, it is necessary to adopt reasonable design methods and optimization methods to reduce the cogging torque of the motor and the vibration noise of the motor during operation. On the other hand, rare earth materials are expensive, and the cost of rare earth permanent magnet synchronous motor is more expensive than that of general electric excitation motor. When designing, it is necessary to compare performance and price according to specific use occasions and requirements. It is also necessary to ensure the magnetic permeability, lightweight and high utilization rate of permanent magnet when optimizing the structure of permanent magnet. Therefore, while reducing the torque pulsation and cogging torque generated during the operation of the motor and the noise generated by the vibration of the motor, improving the magnetic permeability and utilization rate of permanent magnet is an important research direction in the field of permanent magnet structure optimization design in permanent magnet motor.

[0004] The current research shows that the sinusoidal permanent magnet shape can make the excitation magnetic field generated by the permanent magnet in the permanent magnet synchronous motor closer to the sinusoidal shape, so the sinusoidal permanent magnet has smaller air gap magnetic flux harmonics and cogging torque, but the sinusoidal permanent magnet has problems such as difficult processing and high cost during processing. In order to restore the performance of the sinusoidal permanent magnet, most experts and scholars at home and abroad currently use stepped segmentation to replace the sinusoidal shape, but such stepped segmentation will cause a large decrease in the average torque of the motor and cannot effectively restore and retain the excellent performance of the sinusoidal permanent magnet. Therefore, it is very necessary to construct a permanent magnet shape that overcomes the problem of difficult processing of sinusoidal permanent magnets and can better retain the excellent performance of sinusoidal permanent magnets. Summary of the invention

[0005] In view of the shortcomings of existing permanent magnet synchronous motors, such as the generation of cogging torque and vibration noise by permanent magnet materials due to vibration, high temperature and overload current during operation, which may affect the performance of the motor in severe cases, the present invention provides a segmented sinusoidal permanent magnet optimization method for a permanent magnet synchronous motor, which changes the excitation magnetic field generated by the permanent magnet by changing the shape of the permanent magnet, and further changes the magnetic field distribution in the air gap.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is as follows:

[0007] The present invention provides a segmented sinusoidal permanent magnet optimization method for a permanent magnet synchronous motor, which specifically comprises the following steps:

[0008] S1. Use finite element software to establish a standard surface-mounted permanent magnet synchronous motor model;

[0009] S2. Apply the mathematical method of linear interpolation to the sine curve, change the shape of the surface-mounted permanent magnet into a segmented sinusoidal permanent magnet, replace the permanent magnet in the surface-mounted permanent magnet synchronous motor model with a segmented sinusoidal permanent magnet, and establish a segmented sinusoidal permanent magnet synchronous motor model;

[0010] S3. The position of the point on the sine curve of the segmented sinusoidal permanent magnet is taken as the optimization variable, the cogging torque of the permanent magnet synchronous motor and the vibration noise under the rated operating state of the motor are taken as the optimization targets, and the convolutional neural network agent model is combined with the multi-objective genetic algorithm to find the optimal position of the point on the sine curve to achieve the optimization of the segmented sinusoidal permanent magnet.

[0011] In step S2, the segmented sinusoidal permanent magnet shape is achieved by the following steps:

[0012] S21. In the surface-mounted permanent magnet synchronous motor model, the surface shape of the permanent magnet on the side close to the air gap is changed from an arc to a sine curve, and the amplitude, angular velocity, and starting point of the sine curve are determined by the length from the outer diameter of the rotor to the air gap and the mechanical angle occupied by a permanent magnet, and then the mathematical function expression of the sine curve is determined;

[0013] S22. Use the mathematical method of linear interpolation to select points on the sine curve between the two endpoints of a smooth quarter-period, and use straight lines to connect the two endpoints of the quarter-period sine curve and the selected points in sequence to replace the smooth sine curve. Use mirroring and symmetry methods to make straight line shapes between the points of the other half of the permanent magnet that are symmetrical about the y-axis to form a closed permanent magnet shape, which is a segmented sinusoidal permanent magnet.

[0014] In step S3, the position of the point on the curve in the segmented sinusoidal permanent magnet shape is used as the optimization variable, and the cogging torque of the permanent magnet synchronous motor and the vibration noise under the rated operating state of the motor are used as the optimization targets, including:

[0015] S301. The relationship between the cogging torque of the permanent magnet synchronous motor and the shape of the permanent magnet is calculated using the following formula:

[0016]

[0017] Among them, T cog is the cogging torque, Z is the number of stator slots, l Fe is the effective length of the stator core, R1 is the radius of the stator yoke, R2 is the inner diameter of the stator, G n is the Fourier decomposition coefficient of the square of the specific permeability, is the Fourier decomposition coefficient of the square of the air gap magnetic flux density, r is the number of cogging torques generated by one rotation of the motor, P is the number of pole pairs, n is a positive integer, n = 1, 2, 3..., α is the electrical angle, μ0 is the vacuum permeability;

[0018] S302, the vibration noise of the permanent magnet synchronous motor under rated operation is caused by the electromagnetic force waves generated by the interaction of the air gap harmonic magnetic field when the motor is running. The relationship between the radial electromagnetic force and the shape of the permanent magnet is calculated using the following formula:

[0019]

[0020] Among them, p n (θ, t) is the radial electromagnetic force, b 2 (θ, t) is the air gap magnetic density, θ is the angle, t is a certain moment, and μ0 is the vacuum permeability.

[0021] In step S3, the sampling process of the convolutional neural network proxy model and the formation of the network model include:

[0022] S303, classifying different numbers of selected points of the segmented sinusoidal permanent magnet, grouping the number of points used for straight line connection except the first and last points on a quarter period of the sinusoidal curve; sampling the points on the sinusoidal curve using a method combining sub-region sampling and sequential sampling;

[0023] S304, using the functional relationship between the position of the convolutional neural network proxy model prediction point and the cogging torque and the motor vibration noise, the predicted value of the convolutional neural network proxy model is recorded as y, the actual value corresponding to the sampling point is y′, and the error between the two is recorded as ε, that is:

[0024]

[0025] In step S303, the method of combining sub-area sampling with sequential sampling is to divide the sampling area into sub-areas, fix the number of sampling points and the distance between points in each sub-area, and output the initial sampling points; randomly combine the sampling points according to the different numbers of sampling points on the sine curve, and arrange the combined points in coordinate order, and finally output the sampling points.

[0026] The number of points used for straight line connection includes 7 cases: 2, 3, 4, 5, 6, 7, and 8 points. The horizontal coordinates of the corresponding sampling points are expressed as (x1, x2), (x1, x2, x3), (x1, x2, x3, x4), (x1, x2, x3, x4, x5), (x1, x2, x3, x4, x5, x6), (x1, x2, x3, x4, x5, x6, x7), (x1, x2, x3, x4, x5, x6, x7, x8), and the corresponding number of permanent magnet segments is 3, 4, 5, 6, 7, 8, and 9, respectively.

[0027] The present invention has the following beneficial effects and advantages:

[0028] 1. The present invention provides a segmented sinusoidal permanent magnet optimization method for a permanent magnet synchronous motor. Compared with the traditional surface-mounted permanent magnet synchronous motor, the method can reduce the cogging torque and vibration noise, and at the same time reduce the amount of permanent magnets in the permanent magnet synchronous motor and improve the utilization efficiency of the permanent magnets.

[0029] 2. Compared with the sinusoidal permanent magnet whose shape is completely sinusoidal, the segmented sinusoidal permanent magnet of the present invention is in the form of straight line connection between points, which not only solves the problem of difficult and expensive processing of sinusoidal permanent magnets, but also further retains the effect of the sinusoidal permanent magnet in reducing the motor cogging torque and reducing vibration noise.

[0030] 3. The present invention uses an optimization algorithm that combines a convolutional neural network proxy model with a multi-objective genetic algorithm to further optimize the performance of the motor after optimizing the optimal position of the point. Compared with the traditional surface-mounted permanent magnet synchronous motor, the segmented sinusoidal permanent magnet in the permanent magnet synchronous motor after algorithm optimization can also effectively reduce the motor cogging torque and vibration noise in most frequency bands. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a structural schematic diagram of a permanent magnet synchronous motor with an initial segmented sinusoidal permanent magnet shape according to the present invention;

[0032] Figure 2 It is the overall flow chart of the method of the present invention; the formation of shape, innovation.

[0033] Figure 3This is a diagram of the shape forming process of the segmented sinusoidal permanent magnet of the present invention;

[0034] Figure 4 It is a schematic diagram of the optimization method flow of combining a convolutional neural network with a multi-objective genetic algorithm of the present invention;

[0035] Figure 5A This is a comparison diagram of the cogging torque of the surface-mounted, sinusoidal, optimized 2-point segmented sinusoidal, and optimized 3-point segmented sinusoidal permanent magnet synchronous motor models of the present invention;

[0036] Figure 5B This is a comparison diagram of the cogging torque of the surface-mounted permanent magnet synchronous motor, the sinusoidal permanent magnet synchronous motor, the optimized 4-point segmented sinusoidal permanent magnet synchronous motor, and the optimized 5-point segmented sinusoidal permanent magnet synchronous motor models of the present invention;

[0037] Figure 5C This is a comparison diagram of the cogging torque of the surface-mounted, sinusoidal, optimized 6-point segmented sinusoidal, and optimized 7-point segmented sinusoidal permanent magnet synchronous motor models of the present invention;

[0038] Figure 5D This is a comparison diagram of the cogging torque of the surface-mounted, sinusoidal, and optimized 8-point segmented sinusoidal permanent magnet synchronous motor models of the present invention;

[0039] Fig. 6A This is a comparison diagram of the vibration and noise of the surface-mounted, sinusoidal, optimized 2-point segmented sinusoidal, and optimized 3-point segmented sinusoidal permanent magnet synchronous motor models of the present invention;

[0040] Figure 6B This is a comparison diagram of the vibration noise of the surface-mounted, sinusoidal, optimized 4-point segmented sinusoidal, and optimized 5-point segmented sinusoidal permanent magnet synchronous motor models of the present invention;

[0041] Figure 6C This is a comparison diagram of the vibration and noise of the surface-mounted, sinusoidal, optimized 6-point segmented sinusoidal, and optimized 7-point segmented sinusoidal permanent magnet synchronous motor models of the present invention;

[0042] Fig.6D This is a comparison chart of the vibration and noise levels of the surface-mounted, sinusoidal, and optimized 8-point segmented sinusoidal permanent magnet synchronous motor models of the present invention.

[0043] Among them, 1 is the stator, 2 is the stator slot winding, 3 is the rotor core, 4 is the motor shaft, and 5 is the permanent magnet. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0045] like Figure 1 As shown, the permanent magnet synchronous motor of the present invention comprises a stator 1, a rotor 3, and a motor shaft 4, wherein the stator 1 comprises a stator core 1 and a stator winding 2, and the rotor comprises a rotor core 3 and a permanent magnet 5; the present invention provides a segmented sinusoidal permanent magnet optimization method for a permanent magnet synchronous motor as shown in Figure 2 As shown, the following steps are included:

[0046] S1. Use finite element software to establish a standard surface-mounted permanent magnet synchronous motor model;

[0047] S2. Apply the mathematical method of linear interpolation to the sine curve, change the shape of the surface-mounted permanent magnet into a segmented sinusoidal permanent magnet, replace the permanent magnet in the surface-mounted permanent magnet synchronous motor model with a segmented sinusoidal permanent magnet, and establish a segmented sinusoidal permanent magnet synchronous motor model;

[0048] S3, taking the position of the point on the sine curve of the segmented sinusoidal permanent magnet as the optimization variable, the cogging torque of the permanent magnet synchronous motor and the vibration noise under the rated operating state of the motor as the optimization target, and using the method combining the convolutional neural network agent model with the multi-objective genetic algorithm to find the optimal position of the point on the sine curve (process as shown in Figure 4 As shown), the segmented sinusoidal permanent magnet is optimized.

[0049] The size parameters of the permanent magnet synchronous motor in the present invention are shown in Table 1.

[0050] Table 1 shows the size parameters of the permanent magnet synchronous motor of the present invention.

[0051]

[0052] In step S1, according to the basic parameters of the permanent magnet synchronous motor design in Table 1, a standard surface-mounted permanent magnet synchronous motor simulation model is established using Maxwell finite element software.

[0053] In step S2, the segmented sinusoidal permanent magnet shape is achieved by the following steps:

[0054] S21. In the surface-mounted permanent magnet synchronous motor model, the surface shape of the permanent magnet on the side close to the air gap is changed from an arc to a sine curve, and the amplitude, angular velocity, and starting point of the sine curve are determined by the length from the outer diameter of the rotor to the air gap and the mechanical angle occupied by a permanent magnet, and then the mathematical function expression of the sine curve is determined;

[0055] In this embodiment, the amplitude a of the sinusoidal curve function expression in the segmented sinusoidal permanent magnet is 4 mm, and the angular velocity w is The function expression of the sine curve is y=asin(wt+0.5π).

[0056] S22. Use the mathematical method of linear interpolation to select points on the sine curve between the two endpoints of a smooth quarter-period, and use straight lines to connect the two endpoints of the quarter-period sine curve and the selected points in sequence to replace the smooth sine curve. Use mirroring and symmetry methods to make straight line shapes between the points of the other half of the permanent magnet that are symmetrical about the y-axis to form a closed permanent magnet shape, which is a segmented sinusoidal permanent magnet.

[0057] In step S3, the position of the point on the curve in the segmented sinusoidal permanent magnet shape is used as the optimization variable, and the cogging torque of the permanent magnet synchronous motor and the vibration noise under the rated operating state of the motor are used as the optimization targets, including:

[0058] S301. The relationship between the cogging torque of the permanent magnet synchronous motor and the shape of the permanent magnet is calculated using the following formula:

[0059]

[0060] Among them, T cog is the cogging torque, Z is the number of stator slots, l Fe is the effective length of the stator core, R1 is the radius of the stator yoke, R2 is the inner diameter of the stator, G n is the Fourier decomposition coefficient of the square of the specific permeability, It is the Fourier decomposition coefficient of the square of the air gap flux density, r is the number of tooth torques generated by one rotation of the motor, P is the number of pole pairs, n is a positive integer, n = 1, 2, 3..., α is the electrical angle, and μ0 is the vacuum permeability.

[0061] S302, the vibration noise of the permanent magnet synchronous motor under rated operation is caused by the electromagnetic force waves generated by the interaction of the air gap harmonic magnetic field when the motor is running. The relationship between the radial electromagnetic force and the shape of the permanent magnet is calculated using the following formula:

[0062]

[0063] Among them, p n (θ, t) is the radial electromagnetic force, b 2 (θ, t) is the air gap magnetic density, θ is the angle, t is a certain moment, and μ0 is the vacuum permeability.

[0064] From the above theoretical calculations of the cogging torque and vibration noise of the permanent magnet synchronous motor, it can be seen that the cogging torque of the permanent magnet synchronous motor and the vibration noise during operation are related to the distribution of the air gap magnetic field in the permanent magnet synchronous motor, and the distribution of the magnetic field in the air gap is related to the shape of the permanent magnet. Therefore, designing and further optimizing the shape of the permanent magnet can effectively reduce the cogging torque in the motor and the vibration noise of the permanent magnet synchronous motor during operation.

[0065] The sampling process of the convolutional neural network proxy model and the formation of the network model include:

[0066] S303, classifying different numbers of segmented sinusoidal permanent magnet selection points, grouping the number of points used for straight line connection except the first and last points on a quarter period of the sine curve; sampling the points on the sine curve using a method combining sub-area sampling and sequential sampling.

[0067] S304, using the functional relationship between the position of the convolutional neural network proxy model prediction point and the cogging torque and the motor vibration noise, the predicted value of the convolutional neural network proxy model is recorded as y, the actual value corresponding to the sampling point is y′, and the error between the two is recorded as ε, that is:

[0068]

[0069] In step S303, the method of combining sub-area sampling with sequential sampling is to divide the sampling area into sub-areas, fix the number of sampling points and the distance between points in each sub-area, and output the initial sampling points; randomly combine the sampling points according to the different numbers of sampling points on the sine curve, and arrange the combined points in coordinate order, and finally output the sampling points.

[0070] In this embodiment, sampling is performed on a smooth sine curve between two endpoints of the sine function curve y=asin(wt+0.5π) in the first quarter period, and then the two endpoints of the quarter period sine curve and the selected points are sequentially connected by straight lines to replace the smooth sine curve, and a straight line shape between the points of the other half of the permanent magnet symmetrical about the y axis is made by using mirroring and symmetry methods to form a closed permanent magnet shape. The process of forming a segmented sinusoidal permanent magnet shape is shown in FIG. Figure 3 As shown. In this embodiment, the number of points on the curve is divided into 7 cases, and the horizontal coordinates of the corresponding sampling points are respectively expressed as (x1, x2), (x1, x2, x3), (x1, x2, x3, x4), (x1, x2, x3, x4, x5), (x1, x2, x3, x4, x5), (x1, x2, x3, x4, x5, x6), (x1, x2, x3, x4, x5, x6, x7), (x1, x2, x3, x4, x5, x6, x7, x8), and the corresponding number of permanent magnet segments is 3, 4, 5, 6, 7, 8, and 9, respectively. The number of sampling points corresponding to the above 7 cases is 220, 200, 180, 160, 140, 120, and 100, respectively. After the permanent magnet shapes of the 7 cases after classification, the coordinate positions of the points on the sine curve are used as optimization design variables.

[0071] Finite element simulation software Maxwell and Workbench were used to obtain the cogging torque and vibration noise data of the permanent magnet synchronous motor of these seven categories of hundreds of data. The motor cogging torque data of these sampling points were obtained by Maxwell finite element simulation software. In the Workbench finite element software, an electromagnetic transient analysis model of the motor running state was established to calculate the order electromagnetic force of the motor; then a motor harmonic response analysis model was established, and the calculated order electromagnetic force was loaded onto the modal basis of the structure, and then a vibration response analysis was performed to obtain the vibration response results of the motor; a motor harmonic acoustic analysis model was established using Workbench, and the vibration response results of the motor were mapped to the acoustic model, and the acoustic response calculation was performed to obtain the noise results of the motor under the action of radial electromagnetic force.

[0072] The table of the cogging torque and vibration noise values ​​corresponding to the 7 groups of sample points is input into the convolutional neural network proxy model, and the convolutional neural network proxy model is used to predict the functional relationship between the position of the point and the cogging torque and the motor vibration noise, that is, the error formula in step S304 calculates the error between the two.

[0073] In step s3, the process of the multi-objective genetic algorithm includes the following steps:

[0074] For a decision variable x=(x1, x2…x n ) and multi-objective function f i (x) is the optimization problem to define the objective function, i is a positive integer, the present invention selects the two objects of cogging torque and vibration noise as the optimization target, that is, i = 2, then

[0075] minf(x)=(f1(x),f2(x))

[0076] The objective function in the present invention is a response surface model network function generated by a convolutional neural network proxy model;

[0077] Initialize the population: Each individual in the population represents a specific CNN configuration (e.g., number of convolutional layers, number of filters per layer, learning rate, etc.);

[0078] Compute individuals: train a model using the CNN configuration represented by each individual; test these models and record the performance metric of each model as the objective function value of the individual;

[0079] Multi-objective selection: Apply multi-objective genetic algorithm selection strategies (such as non-dominated sorting and crowding distance) to determine which individuals are retained to the next generation;

[0080] Genetic operation: Perform crossover and mutation operations on the selected individuals to generate new individuals. The new individuals inherit certain characteristics of their parents and may have new configurations;

[0081] Repeat the steps of initializing the population, calculating individuals, and genetic operations until the preset accuracy is reached;

[0082] Output results: Output the Pareto optimal solution set in the final generation of population.

[0083] Table 2 shows the optimal position data of the point after optimization by convolutional neural network and multi-objective genetic algorithm, its tooth slot torque and vibration noise results at a frequency of 2000 Hz.

[0084]

[0085] The optimal solution after optimization by the multi-objective genetic algorithm is compared with the traditional surface-mounted and sinusoidal permanent magnets in the initial state. The cogging torque and noise during operation of the permanent magnet synchronous motor are significantly lower than those of the surface-mounted permanent magnet synchronous motor, and the cogging torque result is close to that of the initial sinusoidal permanent magnet synchronous motor. The vibration noise is lower than that of the initial sinusoidal permanent magnet synchronous motor in most frequency ranges. Figures 5A to 5D and Figures 6A to 6D The optimal solution of the segmented sinusoidal permanent magnet after optimization in 7 cases, namely 2, 3, 4, 5, 6, 7, and 8, is compared with the cogging torque of the traditional surface-mounted and sinusoidal permanent magnet synchronous motors, as well as the vibration and noise during operation.

[0086] Figure 5A The optimal solutions of 2-point segmented sine and 3-point segmented sine permanent magnet synchronous motors are compared with the cogging torque of surface-mounted permanent magnet synchronous motors and sine permanent magnet synchronous motors. After optimization, the amplitude of the cogging torque of the 2-point segmented sine permanent magnet synchronous motor is 746.555mN, and the amplitude of the cogging torque of the 3-point segmented sine permanent magnet synchronous motor is 736.721mN. The cogging torque amplitude of the optimal solution of the 2-point and 3-point segmented sine permanent magnet synchronous motor is very close to that of the sine permanent magnet synchronous motor, and is much smaller than the cogging torque amplitude of the traditional surface-mounted permanent magnet synchronous motor.

[0087] Figure 5B The optimal solutions of 4-point segmented sine and 5-point segmented sine permanent magnet synchronous motors are compared with the cogging torque of surface-mounted permanent magnet synchronous motors and sine permanent magnet synchronous motors. After optimization, the amplitude of the cogging torque of the 4-point segmented sine permanent magnet synchronous motor is 750.219mN, and the amplitude of the cogging torque of the 5-point segmented sine permanent magnet synchronous motor is 712.818mN. The cogging torque amplitude of the optimal solution of the 4-point and 5-point segmented sine permanent magnet synchronous motor is very close to that of the sine permanent magnet synchronous motor, and is much smaller than the cogging torque amplitude of the traditional surface-mounted permanent magnet synchronous motor.

[0088] Figure 5C The optimal solutions of 6-point segmented sine permanent magnet synchronous motors and 7-point segmented sine permanent magnet synchronous motors are compared with the cogging torque of surface-mounted permanent magnet synchronous motors and sine permanent magnet synchronous motors. After optimization, the amplitude of the cogging torque of the 6-point segmented sine permanent magnet synchronous motor is 580.41mN, and the amplitude of the cogging torque of the 7-point segmented sine permanent magnet synchronous motor is 899.561mN. The cogging torque amplitude of the optimal solution of the 6-point segmented sine permanent magnet synchronous motor is slightly smaller than that of the sine permanent magnet synchronous motor, and is much smaller than that of the traditional surface-mounted permanent magnet synchronous motor. The cogging torque amplitude of the optimal solution of the 7-point segmented sine permanent magnet synchronous motor is slightly larger than that of the sine permanent magnet synchronous motor, but is much smaller than that of the traditional surface-mounted permanent magnet synchronous motor.

[0089] Figure 5D The optimal solution of the 8-point segmented sinusoidal permanent magnet synchronous motor is compared with the cogging torque of the surface-mounted permanent magnet synchronous motor and the sinusoidal permanent magnet synchronous motor. After optimization, the amplitude of the cogging torque of the 8-point segmented sinusoidal permanent magnet synchronous motor is 698.327mN. The cogging torque amplitude of the optimal solution of the 8-point segmented sinusoidal permanent magnet synchronous motor is slightly smaller than that of the sinusoidal permanent magnet synchronous motor, and is much smaller than that of the traditional surface-mounted permanent magnet synchronous motor.

[0090] Fig. 6A The optimal solutions of 2-point segmented sinusoidal permanent magnet synchronous motors and 3-point segmented sinusoidal permanent magnet synchronous motors were compared with the vibration noise of surface-mounted permanent magnet synchronous motors and sinusoidal permanent magnet synchronous motors in the frequency range of 0-5000Hz. After optimization, the vibration noise of 2-point and 3-point segmented sinusoidal permanent magnet synchronous motors in most frequency bands is smaller than that of sinusoidal permanent magnet synchronous motors and surface-mounted permanent magnet synchronous motors.

[0091] Figure 6B The optimal solutions of 4-point segmented sinusoidal permanent magnet synchronous motors and 5-point segmented sinusoidal permanent magnet synchronous motors were compared with the vibration noise of surface-mounted permanent magnet synchronous motors and sinusoidal permanent magnet synchronous motors in the frequency range of 0-5000Hz. After optimization, the vibration noise of 4-point and 5-point segmented sinusoidal permanent magnet synchronous motors in most frequency bands is smaller than that of sinusoidal permanent magnet synchronous motors and surface-mounted permanent magnet synchronous motors.

[0092] Figure 6C The optimal solutions of 6-point segmented sine permanent magnet synchronous motors and 7-point segmented sine permanent magnet synchronous motors were compared with the vibration noise of surface mounted permanent magnet synchronous motors and sine permanent magnet synchronous motors in the frequency range of 0-5000Hz. After optimization, the vibration noise of 6-point and 7-point segmented sine permanent magnet synchronous motors in most frequency bands is smaller than that of sine permanent magnet synchronous motors and surface mounted permanent magnet synchronous motors.

[0093] Fig.6D The optimal solution of the 8-point segmented sinusoidal permanent magnet synchronous motor is compared with the vibration noise of the surface-mounted permanent magnet synchronous motor and the sinusoidal permanent magnet synchronous motor in the frequency range of 0-5000Hz. After optimization, the vibration noise of the 8-point segmented sinusoidal permanent magnet synchronous motor is smaller than that of the sinusoidal permanent magnet synchronous motor and the surface-mounted permanent magnet synchronous motor in most frequency bands.

[0094] In summary, the present invention combines a smooth sinusoidal curve with a mathematical method of linear interpolation to form a new shape of a segmented sinusoidal permanent magnet in which several points on the sinusoidal curve are interconnected by straight lines. In the optimization process of the segmented sinusoidal permanent magnet, the cogging torque of the motor and the noise under rated operation are taken as optimization targets, and a sampling method combining random sampling and sequential sampling is adopted to sample the points on the sinusoidal curve. The optimal position of the points on the sinusoidal curve is found by combining a convolutional neural network proxy model with a multi-objective genetic algorithm. The originally smooth sinusoidal curve can be converted into a segmented sinusoidal permanent magnet shape that is easier to process, while retaining the excellent performance of the sinusoidal permanent magnet with a smaller cogging torque. The cogging torque and vibration noise of the permanent magnet synchronous motor of the segmented sinusoidal permanent magnet are further reduced by optimizing the method combining a convolutional neural network proxy model with a multi-objective genetic algorithm.

Claims

1. A method for optimizing a segmented sinusoidal permanent magnet of a permanent magnet synchronous motor, characterized in that The following steps are involved: S1. Use finite element software to establish a standard surface-mounted permanent magnet synchronous motor model; S2. Apply the mathematical method of linear interpolation to the sine curve, change the shape of the surface-mounted permanent magnet into a segmented sinusoidal permanent magnet, replace the permanent magnet in the surface-mounted permanent magnet synchronous motor model with a segmented sinusoidal permanent magnet, and establish a segmented sinusoidal permanent magnet synchronous motor model; S3. The position of the point on the sine curve of the segmented sinusoidal permanent magnet is taken as the optimization variable, the cogging torque of the permanent magnet synchronous motor and the vibration noise under the rated operating state of the motor are taken as the optimization targets, and the convolutional neural network agent model is combined with the multi-objective genetic algorithm to find the optimal position of the point on the sine curve to achieve the optimization of the segmented sinusoidal permanent magnet.

2. The segmented sinusoidal permanent magnet optimization method for a permanent magnet synchronous motor according to claim 1, characterized in that In step S2, the segmented sinusoidal permanent magnet shape is achieved by the following steps: S21. In the surface-mounted permanent magnet synchronous motor model, the surface shape of the permanent magnet on the side close to the air gap is changed from an arc to a sine curve, and the amplitude, angular velocity, and starting point of the sine curve are determined by the length from the outer diameter of the rotor to the air gap and the mechanical angle occupied by a permanent magnet, and then the mathematical function expression of the sine curve is determined; S22. Use the mathematical method of linear interpolation to select points on the sine curve between the two endpoints of a smooth quarter-period, and use straight lines to connect the two endpoints of the quarter-period sine curve and the selected points in sequence to replace the smooth sine curve. Use mirroring and symmetry methods to make straight line shapes between the points of the other half of the permanent magnet that are symmetrical about the y-axis to form a closed permanent magnet shape, which is a segmented sinusoidal permanent magnet.

3. The segmented sinusoidal permanent magnet optimization method for a permanent magnet synchronous motor according to claim 1, characterized in that In step S3, the position of the point on the curve in the segmented sinusoidal permanent magnet shape is used as the optimization variable, and the cogging torque of the permanent magnet synchronous motor and the vibration noise under the rated operating state of the motor are used as the optimization targets, including: S301. The relationship between the cogging torque of the permanent magnet synchronous motor and the shape of the permanent magnet is calculated using the following formula: Among them, T cog is the cogging torque, Z is the number of stator slots, l Fe is the effective length of the stator core, R1 is the radius of the stator yoke, R2 is the inner diameter of the stator, G n is the Fourier decomposition coefficient of the square of the specific permeability, is the Fourier decomposition coefficient of the square of the air gap magnetic flux density, r is the number of cogging torques generated by one rotation of the motor, P is the number of pole pairs, n is a positive integer, n = 1, 2, 3..., α is the electrical angle, μ0 is the vacuum permeability; S302, the vibration noise of the permanent magnet synchronous motor under rated operation is caused by the electromagnetic force waves generated by the interaction of the air gap harmonic magnetic field when the motor is running. The relationship between the radial electromagnetic force and the shape of the permanent magnet is calculated using the following formula: Among them, p n (θ, t) is the radial electromagnetic force, b 2 (θ, t) is the air gap magnetic density, θ is the angle, t is a certain moment, and μ0 is the vacuum permeability.

4. The segmented sinusoidal permanent magnet optimization method for a permanent magnet synchronous motor according to claim 1, characterized in that In step S3, the sampling process of the convolutional neural network proxy model and the formation of the network model include: S303, classifying different numbers of selected points of the segmented sinusoidal permanent magnet, grouping the number of points used for straight line connection except the first and last points on a quarter period of the sinusoidal curve; sampling the points on the sinusoidal curve using a method combining sub-region sampling and sequential sampling; S304, using the functional relationship between the position of the convolutional neural network proxy model prediction point and the cogging torque and the motor vibration noise, the predicted value of the convolutional neural network proxy model is recorded as y, the actual value corresponding to the sampling point is y′, and the error between the two is recorded as ε, that is:

5. The method for optimizing the segmented sinusoidal permanent magnet of a permanent magnet synchronous motor according to claim 4, characterized in that In step S303, the method of combining sub-area sampling with sequential sampling is to divide the sampling area into sub-areas, fix the number of sampling points and the distance between points in each sub-area, and output the initial sampling points; randomly combine the sampling points according to the different numbers of sampling points on the sine curve, and arrange the combined points in coordinate order, and finally output the sampling points.

6. The method for optimizing the segmented sinusoidal permanent magnet of a permanent magnet synchronous motor according to claim 5, characterized in that: The number of points used for straight line connection includes 7 cases: 2, 3, 4, 5, 6, 7, and 8 points. The horizontal coordinates of the corresponding sampling points are expressed as (x1, x2), (x1, x2, x3), (x1, x2, x3, x4), (x1, x2, x3, x4, x5), (x1, x2, x3, x4, x5, x6), (x1, x2, x3, x4, x5, x6, x7), (x1, x2, x3, x4, x5, x6, x7, x8), and the corresponding number of permanent magnet segments is 3, 4, 5, 6, 7, 8, and 9, respectively.

Images