Motor stator core orthotropic material parameter identification method based on proxy model

Abstract

The invention relates to a motor stator core orthotropic material parameter identification method based on a proxy model, and the method comprises the steps: extracting effective feature parameters of intrinsic mode features of a motor stator core, establishing a structure-material unification model of the intrinsic mode characteristics of the stator core, and constructing a parameterized finite element model of the stator core; constructing an agent model between the intrinsic mode characteristics of the stator core and the structural parameters and the material parameters; performing a modal test on a batch of to-be-measured sub-cores with a set number to extract the vibration mode of the structure and the inherent frequency corresponding to the vibration mode; and taking the vibration mode of the stator core structure and the inherent frequency corresponding to the vibration mode as target variables, performing iterative optimization on the material parameters of each stator core in combination with the constructed agent model, and obtaining the orthotropic material parameters of the batch of stator cores after iteration is ended. Compared with the prior art, material parameter identification can be efficiently and accurately carried out on a plurality of stator cores of any structures.

Description

Technical Field

[0001] The invention relates to the technical field of modal analysis of motor stator cores, and in particular to a method for identifying orthotropic material parameters of motor stator cores based on a proxy model. Background Art

[0002] The vibration and noise problems of motors during operation have caused many problems for users, and radial electromagnetic force waves and structural modes are two important factors that cause motor vibration and noise. Electromagnetic force waves mainly act on the stator teeth, and then act on the housing through the stator yoke. Therefore, the structural modal analysis of the stator core is of great significance for the analysis and suppression of motor vibration and noise.

[0003] Finite element simulation is a common method for modal analysis of stator core, but the laminated structure of the stator core makes it orthotropic, so accurate orthotropic material parameters are the key to model establishment.

[0004] In the prior art, Chinese patent CN109543282A proposes a method for measuring anisotropic parameters of laminated cores based on modal analysis. This method performs comprehensive and accurate modal measurements on the laminated core structure, and then converts the laminated core structure into a finite element model with an internal toothed surface entity, and identifies material parameters by manually modifying and optimizing the sensitive parameter group of the finite element model.

[0005] Similar methods were used to identify the orthotropic material parameters of the stator core in the literature "Identification of material parameters of rail traction motor core [J]. Journal of Railway Science and Engineering, 2021." and the literature "Modal analysis of claw-pole generator considering anisotropy of stator core and winding [J]. Vibration and Shock, 2017."

[0006] However, the above method has the following disadvantages:

[0007] 1. When identifying the material parameters of a single stator core, this method requires a large amount of finite element modal simulation during parameter adjustment and sensitivity analysis of the finite element model, which is relatively time-consuming, has a large amount of calculation and a high calculation time cost;

[0008] 2. This manual parameter adjustment method has a certain degree of randomness and blindness, so the accuracy of parameter identification cannot be guaranteed;

[0009] 3. This method is not suitable for material parameter identification of stator cores with different structures, because it requires several simulation analyses for each stator core, which takes a lot of time and is therefore inefficient. Summary of the invention

[0010] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a method for identifying orthogonal anisotropic material parameters of motor stator core based on a proxy model, which can efficiently and accurately identify material parameters for a certain number of stator cores with arbitrary structures.

[0011] The object of the present invention can be achieved by the following technical solution: A method for identifying orthotropic material parameters of a motor stator core based on a proxy model, comprising the following steps:

[0012] S1. Extracting effective characteristic parameters of the inherent modal characteristics of the motor stator core, specifically including: effective structural parameters and effective material parameters;

[0013] S2. Using the extracted effective characteristic parameters, a unified structure-material model of the inherent modal characteristics of the stator core is established;

[0014] And based on this, a parametric finite element model of the stator core is constructed;

[0015] S3, constructing a proxy model between the inherent modal characteristics of the stator core and the structural parameters and material parameters;

[0016] S4, performing modal tests on a set number of sub-cores to be tested to extract the vibration mode of the structure and its corresponding natural frequency;

[0017] S5. The vibration mode of the stator core structure and its corresponding natural frequency are used as target variables. Combined with the constructed proxy model, the material parameters of each stator core are iteratively optimized. After the iteration is terminated, the orthogonal anisotropic material parameters of this batch of stator cores are obtained.

[0018] Furthermore, the effective structural parameters and effective material parameters of the stator core natural modal characteristics in step S1 specifically include: R i ,σ,λ,δ,τ,E x ,E z ,G xy ,G xz , respectively represent the inner radius of the stator core, the crack ratio (the ratio of the inner radius to the outer radius of the core), the aspect ratio (the ratio of the axial length of the core to the outer diameter), the tooth height ratio (the ratio of the tooth height to the core height), the tooth slot ratio (the ratio of the tooth width to the slot width), the elastic modulus of the stator core in the x direction, the elastic modulus of the stator core in the z direction (axial direction), the shear modulus of the stator core in the xy plane, and the shear modulus of the stator core in the xz plane.

[0019] Furthermore, the structure-material unified model in step S2 is a conceptual model for characterizing the relationship between the inherent modal characteristics of the stator core and its structural parameters and material parameters:

[0020] F=f(R i ,σ,λ,δ,τ,E x ,E z ,G xy ,G xz )

[0021] Wherein, F is the natural modal characteristic of the stator core, and the parameters on the right side of the equation are characteristic parameters related to the natural modal characteristic of the stator core.

[0022] Furthermore, the step S3 specifically includes the following steps:

[0023] S31, prepare the data set required for training the proxy model;

[0024] S32, selecting 90% of the obtained data set capacity as a training set, and using a convolutional neural network to train a proxy model that can reflect the nonlinear relationship between the effective structural parameters and effective material parameters of the stator core and its inherent modal characteristics, wherein the input of the proxy model is: the effective structural parameters and effective material parameters of the stator core, and the output is: the inherent modal characteristics of the stator core;

[0025] S33. 10% of the obtained data set capacity is selected as a validation set, and the accuracy of the proxy model is verified using an error function.

[0026] Furthermore, the step S31 specifically includes the following steps:

[0027] S311, using the effective feature parameters extracted in step S1 as independent variables, and using the Latin hypercube sampling method to perform experimental design of the data set;

[0028] S312, using the parameterized finite element model of the stator core constructed in step S2, calculating the natural modal characteristics of each sample in the data set, specifically including 6 low-order modal characteristics: second-order same direction, second-order opposite direction, third-order same direction, third-order opposite direction, fourth-order same direction and fourth-order opposite direction vibration modes and their corresponding natural frequencies;

[0029] S313. Combine the effective characteristic parameters and low-order modal features of each sample to obtain a complete data set.

[0030] Furthermore, the error function in step S33 is specifically:

[0031]

[0032] Among them, E r is the relative error, F SM,i is the natural frequency corresponding to the i-th vibration mode predicted by the surrogate model, F FEM,i is the natural frequency corresponding to the i-th vibration mode obtained by finite element simulation.

[0033] Furthermore, the vibration modes of the structure and their corresponding natural frequencies extracted in step S4 are specifically: second-order same-direction, second-order opposite-direction, third-order same-direction, third-order opposite-direction, fourth-order same-direction and fourth-order opposite-direction vibration modes and their corresponding natural frequencies.

[0034] Furthermore, the step S5 specifically includes the following steps:

[0035] S51. Using the trained proxy model to replace the finite element modal analysis required for iterative optimization in the traditional method, and using the proxy model to predict the low-order natural modal characteristics of the stator core with any structure and any material parameters;

[0036] S52. A parallel particle swarm optimization algorithm is used to construct a fitness function and impose constraints to simultaneously identify material parameters of several stator cores, and finally obtain material parameters that minimize the fitness function value, which are the required orthogonal anisotropic material parameters of each stator core.

[0037] Furthermore, the fitness function in step S52 is specifically:

[0038]

[0039] Among them, Err i represents the fitness of the ith population. The value of N is 6 because the main focus is on the six low-order modal characteristics of the stator core. SM,ij represents the natural frequency corresponding to the jth vibration mode of the individual in the i-th population obtained by the surrogate model, F Test,ij It represents the natural frequency corresponding to the jth vibration mode of the i-th stator core obtained through modal test.

[0040] Furthermore, the constraint conditions in step S52 are specifically:

[0041]

[0042] in, σ m ,λ m ,δ m ,τ m They respectively represent the measured values ​​of the corresponding structural parameters of each stator core, and the value ranges of the four effective material parameters are determined through literature research and design experience.

[0043] Compared with the prior art, the present invention has the following advantages:

[0044] The present invention extracts effective characteristic parameters (including effective structural parameters and effective material parameters) of the inherent modal characteristics of the motor stator core to establish a unified structure-material model of the inherent modal characteristics of the stator core, and thereby constructs a parametric finite element model of the stator core; further constructs a proxy model between the inherent modal characteristics of the stator core and the structural parameters and material parameters; uses the constructed modal characteristic proxy model to replace the finite element modal analysis of the stator core, and based on the proxy model, realizes the process of material parameter identification through algorithm optimization, which replaces a large number of manual parameter adjustments and simulation attempts in traditional methods, and is conducive to improving the efficiency and accuracy of material parameter identification.

[0045] The present invention takes into account both structural parameters and material parameters to construct a modal proxy model to replace finite element simulation parameter adjustment, which makes the present invention universally applicable and can realize the synchronous identification of several different stator core material parameters: not only can the material parameters of stator cores of various structures be identified, but also there is no need to retrain the model and update the data set for new individuals, and the material parameters of a certain number of stator cores can be identified in batches, which greatly improves the application efficiency of the present invention and realizes wide applicability across individuals. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0047] Figure 2 A schematic diagram of the original dimensions involved in the definition of effective structural parameters;

[0048] Figure 3 A schematic diagram of a finite element model established in the embodiment and a coordinate system used;

[0049] Figure 4a Schematic diagram of error distribution of the 24-slot stator core proxy model constructed in the embodiment for the prediction of the validation set;

[0050] Figure 4b Schematic diagram of error distribution of the 36-slot stator core proxy model constructed in the embodiment for the prediction of the validation set;

[0051] Figure 4c Schematic diagram of error distribution of the 48-slot stator core proxy model constructed in the embodiment for the prediction of the validation set;

[0052] Figure 5a Schematic diagram of the arrangement of vibration pickup points and excitation points for the modal test of prototype 1 in the embodiment;

[0053] Figure 5b Schematic diagram of the arrangement of vibration pickup points and excitation points for the modal test of prototype 2 in the embodiment;

[0054] Figure 5cSchematic diagram of the arrangement of vibration pickup points and excitation points for the modal test of prototype 3 in the embodiment;

[0055] Figure 5d Schematic diagram of the arrangement of the vibration pickup points and the excitation points of the modal test of the prototype 4 in the embodiment;

[0056] Figure 6 Schematic diagram of the process of the parallel particle swarm optimization algorithm used in the embodiment;

[0057] Figure 7 Schematic diagram of the evolution process of the parallel particle swarm optimization algorithm when performing material parameter identification of the prototype in the embodiment. DETAILED DESCRIPTION

[0058] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0059] Example

[0060] like Figure 1 As shown, a method for identifying parameters of orthotropic material of motor stator core based on surrogate model comprises the following steps:

[0061] S1. Extracting effective characteristic parameters of the inherent modal characteristics of the motor stator core, specifically including: effective structural parameters and effective material parameters;

[0062] S2. Using the extracted effective characteristic parameters, a unified structure-material model of the inherent modal characteristics of the stator core is established;

[0063] And based on this, a parametric finite element model of the stator core is constructed;

[0064] S3, constructing a proxy model between the inherent modal characteristics of the stator core and the structural parameters and material parameters;

[0065] S4, performing modal tests on a set number of sub-cores to be tested to extract the vibration mode of the structure and its corresponding natural frequency;

[0066] S5. The vibration mode of the stator core structure and its corresponding natural frequency are used as target variables. Combined with the constructed proxy model, the material parameters of each stator core are iteratively optimized. After the iteration is terminated, the orthogonal anisotropic material parameters of this batch of stator cores are obtained.

[0067] The above method is applied to this embodiment. In the embodiment, four drive motor stator cores of different specifications are selected to form a group of test pieces. Their specific parameters are given in Table 1. The embodiment is used to further explain the present invention in detail, which is a further explanation of the present invention rather than a limitation.

[0068] Table 1 Specific parameters of each prototype

[0069] parameter Prototype 1 Prototype 2 Prototype 3 Prototype 4 Pole pairs 4 2 4 4 Number of slots 24 36 48 48 Slot width 2.4mm 3mm 2mm 2mm Air gap length 1mm 0.3mm 1mm 0.75mm Stator outer diameter 140mm 155mm 180mm 220mm Rotor outer diameter 86mm 97.4mm 125mm 142mm Axial length 84mm 105mm 96mm 156mm

[0070] The main application processes are:

[0071] Step 1: extracting effective characteristic parameters of the inherent modal characteristics of the motor stator core, specifically including: effective structural parameters and effective material parameters.

[0072] This scheme takes into account that the inherent modal characteristics of the multi-degree-of-freedom system are closely related to the two parameters of mass and stiffness. The structural characteristics mainly affect the mass and distribution of the core, so the extracted structural characteristic parameters should be able to basically determine the three-dimensional structure of a stator core. Based on this principle, the extracted effective structural parameters are as follows: R i ,σ,λ,δ,τ, respectively represent the inner radius, split ratio, aspect ratio, tooth height ratio and slot ratio of the stator core, where the split ratio is the ratio of the inner radius to the outer radius of the core, the aspect ratio is the ratio of its axial length to the outer diameter, the tooth height ratio is the ratio of the tooth height to the core height, and the slot ratio is the ratio of the tooth width to the slot width. There are a total of 5 effective structural parameters. The original dimensions involved in the definition of each parameter are as follows: Figure 2-3 As shown;

[0073] The material parameters mainly affect the stiffness of the core. Considering that the stator core is stacked along the axial direction, the coordinate system used in this embodiment is as follows: Figure 3 As shown, the Young's modulus E and shear modulus G of the laminated core in each direction have the following relationship: E x =E y ,G xz =G yz , while the Poisson's ratio has little effect on the model characteristics. Usually, the value of the lamination material is taken by default in finite element modeling. In this embodiment, the Poisson's ratio of the stator core model is the value of the Poisson's ratio of the silicon steel sheet, that is, υ xy =υ xz =υ yz =0.29, so the final extracted effective material parameter is E x ,E z ,G xy and G xz , a total of 4 valid material parameters.

[0074] Step 2: Based on the 9 effective characteristic parameters extracted in step 1, a unified conceptual model of the inherent modal characteristics of the stator core with respect to structural parameters and material parameters is established:

[0075] F=f(R i ,σ,λ,δ,τ,E x ,E z ,G xy ,G xz ),

[0076] Wherein, F represents the inherent modal characteristics of the stator core, and the parameters on the right side of the equation are characteristic parameters related to the inherent modal characteristics of the stator core.

[0077] Based on the above conceptual model, this embodiment uses Ansys software to build a model, parameterizes all the effective features extracted in step 1, and establishes a finite element model as shown in Figure 3 shown.

[0078] Step 3: Construct a proxy model between the natural modal characteristics of the stator core and the structural parameters and material parameters. The large number of parameter adjustment attempts based on finite element modal analysis in traditional methods is an important reason for their low efficiency. Therefore, this embodiment uses a proxy model to replace finite element modal simulation, and uses the proxy model to predict the natural frequency of the vibration mode to be extracted, which will effectively improve the efficiency of parameter identification.

[0079] In step 3, the construction of the proxy model for characterizing the relationship between the natural modal characteristics and the effective characteristic parameters of the stator core is performed by the following steps:

[0080] Step 3.1: Use the 9 effective feature parameters extracted in step 1 as independent variables, use the Latin hypercube sampling method to design the data set experiment, and determine the value range of each parameter based on design experience and research. The value range of each parameter is shown in Table 2.

[0081] In this embodiment, data sets are prepared for stator cores with 24, 36 and 48 slots, respectively, and the sample capacity is 200, in order to use the above data sets to accurately predict the natural modal characteristics of stator cores with common slot numbers and arbitrary structures and material parameters.

[0082] Table 2 Upper and lower limits of effective characteristic parameter values ​​in each data set

[0083]

[0084] Step 3.2: Use the parameterized finite element model constructed in step 2 to calculate the inherent modal characteristics of each sample in the data set, and extract the six low-order modal characteristics of each sample as the dependent variables of the data set, including: second-order same-direction, second-order opposite-direction, third-order same-direction, third-order opposite-direction, fourth-order same-direction and fourth-order opposite-direction vibration modes and their corresponding natural frequencies.

[0085] Step 3.3: Use the characteristic parameters of each sample in step 3.1 and the inherent modal features corresponding to the sample calculated in step 3.2 as the input and output of the proxy model training, thereby obtaining a complete data set.

[0086] Step 3.4: Split the obtained data set, randomly select 180 groups of samples as training sets, and the remaining 20 groups of samples as validation sets.

[0087] Step 3.5: In this embodiment, a convolutional neural network is used to train the proxy model to achieve accurate prediction of the natural modal frequency.

[0088] The hyperparameters of the convolutional neural network used in this embodiment are shown in Table 3.

[0089] Table 3 Hyperparameters of the convolutional neural network used in the embodiment

[0090] Hyperparameters value Learning Rate 0.01 Batch size 36 Number of convolutional layers 3 The number of convolution kernels in each convolution layer 16，32，64 Convolution kernel size 3 Number of fully connected layers 2 Number of neurons in each fully connected layer 64，8

[0091] Step 3.6: Verify the accuracy of the proxy model trained in step 3.5, that is, the prediction accuracy of the model for the natural frequency of the corresponding vibration mode of the stator core. The prediction accuracy of the proxy model determines whether it can replace the finite element modal simulation. The error function used in the verification is:

[0092]

[0093] Among them, E r is the relative error, F SM,i is the natural frequency corresponding to the i-th vibration mode predicted by the surrogate model, F FEM,i is the natural frequency corresponding to the i-th vibration mode obtained by finite element simulation.

[0094] The error distribution of the proxy models of the stator core with three common slot numbers trained in this embodiment is as follows: Figure 4a to Figure 4c As shown by Figures 4a to 4c It can be seen that the prediction errors of the proxy models of the stator cores with three slot numbers for the validation set data are all within 5%, which shows that the proxy model can better predict the modal characteristics of the stator core, and thus can replace the finite element simulation in the subsequent parameter identification process.

[0095] Step 4: Perform modal tests on the stator cores of the four prototypes. In this embodiment, a force hammer is used for excitation and an acceleration sensor is used for vibration pickup. In order to clearly distinguish the positive and reverse vibration modes of the same order mode, no less than two circles of vibration pickup points are arranged on the axial direction of the stator core. The vibration pickup points and excitation points of each prototype in the modal test are arranged as follows: Figure 5a to Figure 5d The results of the stator core modal tests of various prototypes are shown in Table 4.

[0096] Table 4 Modal test results of each prototype

[0097]

[0098] Step 5: Use the parallel particle swarm optimization algorithm to Optimize the design of the fitness function and obtain the parameter value that minimizes the fitness function. i represents the fitness of the i-th population. In this embodiment, the maximum value of i is 4 because there are 4 prototypes in total. The value of N is 6 because the main focus is on the 6 low-order modal characteristics of the stator core. SM,ij represents the natural frequency corresponding to the jth vibration mode of the individual in the i-th population obtained by the modal surrogate model, F Test,ij It represents the natural frequency corresponding to the jth vibration mode of the i-th prototype obtained through modal test.

[0099] F SM,ij The value of is obtained through the proxy model, F Test,ij The value of is read from Table 4. The constraints for iterative updating of each parameter in the parallel particle swarm optimization algorithm are as follows:

[0100]

[0101] in, σ m ,λ m ,δ m ,τ m Respectively represent the measured values ​​of the corresponding structural parameters of each stator core. In this embodiment, the effective structural parameters of prototypes 1 to 4 are:

[0102] [44,0.63,0.6,0.65,0.86], [49,0.63,0.68,0.51,0.69], [63.50,0.71,0.53,0.51,1.08], [71.75,0.65,0.71,0.58,1.39].

[0103] The process of the parallel particle swarm optimization algorithm used in this embodiment is as follows: Figure 6 As shown, there are a total of 4 populations, each of which independently handles the material parameter optimization problem of a prototype, and finally completes the material parameter identification of the stator cores of this batch of prototypes at one time.

[0104] The parameter identification in this embodiment is carried out on NVIDIA GeForce RTX 3060. The material parameter identification of the stator cores of the four prototypes is carried out simultaneously based on the parallel particle swarm optimization algorithm. The evolution process of the algorithm is as follows: Figure 7 As shown, the algorithm can basically converge in about 80 generations, and the entire parameter identification process only consumes about 60 seconds. This shows that the method proposed in the present invention is highly efficient in identifying a certain number of material parameters of the stator core.

[0105] The identified anisotropic material parameters of the stator cores of the various prototypes are shown in Table 5. In order to verify the feasibility of the method used in this embodiment, the anisotropic material parameters of the various prototypes obtained in this embodiment are imported into the finite element model for simulation, and the comparison results of the natural frequencies in the simulation and modal test are shown in Table 6.

[0106] Table 5 Anisotropic material parameters of stator core of each prototype in the embodiment

[0107] Material parameters Prototype 1 Prototype 2 Prototype 3 Prototype 4 <![CDATA[ρ / (kg·m -3 )]]> 7303 7296 7378.5 7415 <![CDATA[E X (E Y ) / GPa]]> 190.63 138.95 141.86 208.35 <![CDATA[E Z / GPa]]> 89.23 95.75 116.55 137.42 <![CDATA[G XY / GPa]]> 65.91 63.48 74.42 69.97 <![CDATA[G XZ (G YZ ) / GPa]]> 19.75 10.50 6.22 7.33 <![CDATA[υ XY (u XZ ,v YZ )]]> 0.29 0.29 0.29 0.29

[0108] Table 6 Anisotropic material parameters of stator cores of various prototypes in the embodiments

[0109]

[0110] As shown in Table 6, the maximum error between the simulation and the modal test results of each prototype machine based on the material parameters identified by the method proposed in the present invention does not exceed 3%. Among them, the relative error between the simulation and the test of the third-order forward modal feature is the largest, which is 2.7%, while the relative errors of the other modal features do not exceed 2%. This verifies the accuracy of the proposed method of using a proxy model instead of finite element simulation to perform parameter optimization to identify the material parameters of the stator core.

[0111] In summary, this scheme uses the constructed modal feature proxy model to replace the finite element modal analysis of the stator core, and realizes the material parameter identification process through algorithm optimization based on the proxy model. Compared with the large number of manual parameter adjustments and simulation attempts in traditional methods, this scheme can effectively improve the efficiency and accuracy of material parameter identification.

[0112] Since the effective feature parameters used in constructing the proxy model include both structural parameters and material parameters, this scheme has universal applicability: that is, it can identify the material parameters of stator cores of various structures without the need to retrain the model and update the data set for new individuals, thus achieving wide applicability across individuals.

[0113] Since the scheme has universal applicability and can realize synchronous identification of multiple different stator core material parameters, the material parameters of a certain number of stator cores can be identified in batches, which greatly improves the practical application efficiency of the scheme.

Claims

1. A method for identifying orthotropic material parameters of motor stator core based on a proxy model, characterized in that: The following steps are involved: S1. Extracting effective characteristic parameters of the inherent modal characteristics of the motor stator core, specifically including: effective structural parameters and effective material parameters; S2. Using the extracted effective characteristic parameters, a unified structure-material model of the inherent modal characteristics of the stator core is established; And based on this, a parametric finite element model of the stator core is constructed; S3, constructing a proxy model between the inherent modal characteristics of the stator core and the structural parameters and material parameters; S4, performing modal tests on a set number of sub-cores to be tested to extract the vibration mode of the structure and its corresponding natural frequency; S5. The vibration mode of the stator core structure and its corresponding natural frequency are used as target variables. Combined with the constructed proxy model, the material parameters of each stator core are iteratively optimized. After the iteration is terminated, the orthogonal anisotropic material parameters of this batch of stator cores are obtained.

2. The method for identifying orthotropic material parameters of a motor stator core based on a proxy model according to claim 1, characterized in that: The effective structural parameters and effective material parameters of the stator core natural modal characteristics in step S1 specifically include: i ,σ,λ,δ,τ,E x ,E z ,G xy ,G xz , respectively represent the inner radius of the stator core, the crack ratio, that is, the ratio of the inner radius to the outer radius of the core, the aspect ratio, that is, the ratio of the axial length of the core to the outer diameter, the tooth height ratio, that is, the ratio of the tooth height to the core height, the tooth slot ratio, that is, the ratio of the tooth width to the slot width, the elastic modulus of the stator core in the x direction, the elastic modulus of the stator core in the z direction, that is, the axial direction, the shear modulus of the stator core in the xy plane and the shear modulus of the stator core in the xz plane.

3. The method for identifying orthotropic material parameters of a motor stator core based on a proxy model according to claim 2, characterized in that: The structure-material unified model in step S2 is a conceptual model for characterizing the relationship between the inherent modal characteristics of the stator core and its structural parameters and material parameters: F=f(R i ,s,l,d,t,E x ,E z ,G xy ,G xz ) Wherein, F is the natural modal characteristic of the stator core, and the parameters on the right side of the equation are characteristic parameters related to the natural modal characteristic of the stator core.

4. The method for identifying orthotropic material parameters of a motor stator core based on a proxy model according to claim 2, characterized in that: The step S3 specifically comprises the following steps: S31, prepare the data set required for training the proxy model; S32, selecting 90% of the obtained data set capacity as a training set, and using a convolutional neural network to train a proxy model that can reflect the nonlinear relationship between the effective structural parameters and effective material parameters of the stator core and its inherent modal characteristics, wherein the input of the proxy model is: the effective structural parameters and effective material parameters of the stator core, and the output is: the inherent modal characteristics of the stator core; S33. 10% of the obtained data set capacity is selected as a validation set, and the accuracy of the proxy model is verified using an error function.

5. The method for identifying parameters of orthotropic material of motor stator core based on proxy model according to claim 4, characterized in that: The step S31 specifically includes the following steps: S311, using the effective feature parameters extracted in step S1 as independent variables, and using the Latin hypercube sampling method to perform experimental design of the data set; S312, using the parameterized finite element model of the stator core constructed in step S2, calculating the natural modal characteristics of each sample in the data set, specifically including 6 low-order modal characteristics: second-order same direction, second-order opposite direction, third-order same direction, third-order opposite direction, fourth-order same direction and fourth-order opposite direction vibration modes and their corresponding natural frequencies; S313. Combine the effective characteristic parameters and low-order modal features of each sample to obtain a complete data set.

6. The method for identifying orthotropic material parameters of a motor stator core based on a proxy model according to claim 5, characterized in that: The error function in step S33 is specifically: Among them, E r is the relative error, F SM,i is the natural frequency corresponding to the i-th vibration mode predicted by the surrogate model, F FEM,i is the natural frequency corresponding to the i-th vibration mode obtained by finite element simulation.

7. The method for identifying orthotropic material parameters of a motor stator core based on a proxy model according to claim 1, characterized in that: The vibration modes of the structure and their corresponding natural frequencies extracted in step S4 are specifically: second-order same-direction, second-order opposite-direction, third-order same-direction, third-order opposite-direction, fourth-order same-direction and fourth-order opposite-direction vibration modes and their corresponding natural frequencies.

8. The method for identifying parameters of orthotropic material of motor stator core based on proxy model according to claim 2, characterized in that: The step S5 specifically comprises the following steps: S51. Using the trained proxy model to replace the finite element modal analysis required for iterative optimization in the traditional method, and using the proxy model to predict the low-order natural modal characteristics of the stator core with any structure and any material parameters; S52. A parallel particle swarm optimization algorithm is used to construct a fitness function and impose constraints to simultaneously identify material parameters of several stator cores, and finally obtain material parameters that minimize the fitness function value, which are the required orthogonal anisotropic material parameters of each stator core.

9. The method for identifying parameters of orthotropic material of motor stator core based on proxy model according to claim 8, characterized in that: The fitness function in step S52 is specifically: Among them, Err i represents the fitness of the ith population. The value of N is 6 because the main focus is on the six low-order modal characteristics of the stator core. SM,ij represents the natural frequency corresponding to the jth vibration mode of the individual in the i-th population obtained by the surrogate model, F Test,ij It represents the natural frequency corresponding to the jth vibration mode of the i-th stator core obtained through modal test.

10. The method for identifying parameters of orthotropic material of motor stator core based on proxy model according to claim 9, characterized in that: The constraint conditions in step S52 are specifically: in, σ m ,λ m ,δ m ,τ m They respectively represent the measured values ​​of the corresponding structural parameters of each stator core, and the value ranges of the four effective material parameters are determined through literature research and design experience.

Images