Driving motor NVH performance evaluation method and device
By performing spatial domain equivalent synthesis and frequency domain-order joint analysis of the electromagnetic force in the slanted pole section of the drive motor, and combining it with a machine learning model, the computational complexity of the full-order multi-physics coupled simulation model was solved, enabling rapid and accurate NVH performance evaluation and optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHENZHI AUTOMOBILE TECHNOLOGY GROUP CO LTD CHONGQING INNOVATION RESEARCH BRANCH
- Filing Date
- 2026-01-26
- Publication Date
- 2026-05-08
AI Technical Summary
Existing full-order multiphysics coupling simulation models have high computational complexity, high resource consumption, poor real-time performance, are difficult to integrate, and have low optimization efficiency, failing to meet the needs of rapid feedback and system-level simulation in the early stages of motor design.
Electromagnetic simulation was used to obtain time-domain electromagnetic force distribution data of the skewed pole section of the motor. Spatial domain equivalent synthesis and frequency domain-order joint analysis were performed to extract target frequency points, principal order and phase difference characterization quantities. A lightweight NVH performance evaluation model was constructed and a machine learning model was used for prediction.
It achieves millisecond-level NVH performance prediction on ordinary computing devices, maintains high accuracy, reduces computing resource consumption, is suitable for whole vehicle system-level simulation and online monitoring, and supports rapid parametric research and optimization.
Smart Images
Figure CN121996964A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of motor technology, and specifically to a method and apparatus for evaluating the NVH performance of a drive motor. Background Technology
[0002] With the rapid development of the electric vehicle industry, the performance requirements of drive systems are constantly increasing. Among them, NVH (noise, vibration, and harshness) performance has become a key indicator for measuring the quality of drive motors and directly affecting the overall vehicle ride comfort. Accurately predicting and effectively suppressing NVH problems of drive motors is of great significance for improving the acoustic quality of the entire vehicle and enhancing the market competitiveness of products.
[0003] In current research on NVH (Noise, Vibration, and Harshness) of drive motors, the mainstream technical approach is to construct high-fidelity multiphysics coupled simulation models (typically covering electromagnetic fields, structural dynamic fields, and acoustic fields) based on physical mechanisms. Such full-order models can comprehensively depict the entire process of electromagnetic excitation force generation, structural dynamic response transmission, and radiated noise generation and propagation within the motor, possessing high analytical accuracy and engineering reference value.
[0004] However, the full-order multiphysics coupling model has significant limitations, mainly in the following three aspects:
[0005] 1. High computational complexity and resource consumption: Full-order models typically contain a massive number of degrees of freedom. When performing transient dynamic simulations or frequency response analyses, high-performance computing clusters are required, and a single simulation can take several hours or even days. In scenarios where frequent parametric studies, scheme comparisons, and multiple rounds of iterative optimization are needed in the early stages of motor design, its efficiency is difficult to meet the requirements for rapid feedback in engineering practice.
[0006] 2. Poor real-time performance and difficulty in system integration: Due to the large scale of the model and the complexity of the solution process, it is difficult to embed it into system-level simulation platforms, such as the joint simulation of vehicle multibody dynamics or electric drive system. It also cannot support online NVH assessment or the development of active noise control strategies, which limits its deployment capability in a wider range of engineering applications.
[0007] 3. Low optimization efficiency: In the design optimization process for NVH performance, parameters such as skew pole angle, slot pole matching, and stator structure need to be fine-tuned, requiring repeated evaluation of a large number of design schemes. The high computational cost of full-order models makes simulation-based optimization methods extremely time-consuming and uneconomical, severely restricting the realization of an efficient design process. Summary of the Invention
[0008] The purpose of this invention is to provide a method and apparatus for switching modulation modes of a motor controller, which can solve the problems of high computational cost, low efficiency and difficulty in integration of existing full-order multi-physics coupling simulation methods.
[0009] To achieve the above objectives, the technical solution adopted by the present invention is as follows:
[0010] In a first aspect, the present invention discloses a method for evaluating the NVH performance of a drive motor, which includes the following steps:
[0011] S1. The time-domain electromagnetic force distribution data on the stator teeth of each skew pole section of the target motor is obtained through electromagnetic simulation. The time-domain electromagnetic force distribution data includes radial and tangential components.
[0012] S2, perform spatial domain equivalent synthesis of the electromagnetic forces of each skew pole segment to generate a global equivalent electromagnetic force time-domain signal characterizing the overall excitation of the motor;
[0013] S3, perform frequency domain-order joint analysis on the global equivalent electromagnetic force time domain signal, identify the target frequency point that has a significant impact on NVH performance, and extract the main spatial order and main time order corresponding to the target frequency point to obtain the synthetic excitation force amplitude corresponding to the main spatial order and main time order.
[0014] S4, based on the time-domain electromagnetic force distribution data described in S1, at the target frequency point and under the main spatial order and main time order, calculate the electromagnetic force phase difference between at least two sets of slant pole segments, and generate a phase difference characterization quantity.
[0015] S5, input the target frequency point, the amplitude of the synthetic excitation force and the phase difference characterization quantity into the pre-constructed NVH performance evaluation model, and output the corresponding NVH performance index.
[0016] Furthermore, the NVH performance indicators include acoustic power or equivalent radiated acoustic power.
[0017] Furthermore, the NVH performance evaluation model is obtained by training with multi-physics coupled simulation data or measured test data and is constructed using a machine learning model; the machine learning model includes a support vector machine model, a neural network model, or a Gaussian process regression model.
[0018] Furthermore, the neural network model is a BP neural network model; the input layer of the BP neural network model includes three neurons, which are used to receive the target frequency point, the amplitude of the synthesized excitation force, and the phase difference representation quantity, respectively; the output layer of the BP neural network model includes one neuron, which is used to output NVH performance indicators; the BP neural network model adopts a nonlinear activation function and is trained through a numerical optimization algorithm.
[0019] Furthermore, the frequency domain-order joint analysis is implemented using a two-dimensional Fourier transform, which is used to simultaneously analyze the harmonic characteristics of electromagnetic excitation in the time and space dimensions.
[0020] Secondly, this invention discloses a drive motor NVH performance evaluation device, which includes:
[0021] The electromagnetic force data acquisition module is used to perform the following: acquiring time-domain electromagnetic force distribution data on the stator teeth of each skew pole section of the target motor through electromagnetic simulation, wherein the time-domain electromagnetic force distribution data includes radial and tangential components;
[0022] The excitation synthesis module is used to perform: spatial domain equivalent synthesis of the electromagnetic force of each skewed pole segment, and generate a global equivalent electromagnetic force time-domain signal characterizing the overall excitation of the motor;
[0023] The main order feature extraction module is used to perform: frequency domain-order joint analysis on the global equivalent electromagnetic force time domain signal, identify the target frequency point that has a significant impact on NVH performance, and extract the main spatial order and main time order corresponding to the target frequency point to obtain the synthetic excitation force amplitude corresponding to the main spatial order and main time order;
[0024] The phase difference calculation module is used to perform the following: based on the time-domain electromagnetic force distribution data acquired by the electromagnetic force data acquisition module, at the target frequency point and the main spatial order and the main time order, calculate the electromagnetic force phase difference between at least two sets of slant pole segments, and generate a phase difference characterization quantity.
[0025] The evaluation model module is used to perform the following: inputting the target frequency point, the amplitude of the synthetic excitation force, and the phase difference characterization quantity into the pre-built NVH performance evaluation model, and outputting the corresponding NVH performance indicators.
[0026] The present invention has the following unexpected beneficial effects:
[0027] 1. This invention performs spatial domain equivalent synthesis of the electromagnetic forces in each skewed pole segment and combines frequency domain-order joint analysis to extract the target frequency point, the amplitude of the synthesized excitation force corresponding to the principal order, and the phase difference characterization between the skewed pole segments as low-dimensional input features, avoiding direct solution of the high-degree-of-freedom full-order model. The NVH performance evaluation model constructed in this way can achieve millisecond-level prediction on ordinary computing devices, significantly shortening the single evaluation time and meeting the engineering needs of rapid parameterization research and multi-scheme comparison in the early stage of motor design. Furthermore, the core physical information affecting NVH performance is preserved during the order reduction process, including the vector characteristics of the electromagnetic force, the excitation amplitude of the dominant spatiotemporal order, and the phase relationship between the skewed pole segments. Thus, while significantly reducing the consumption of computing resources, it can still maintain prediction accuracy comparable to full-order multiphysics simulation, achieving an effective balance between efficiency and accuracy.
[0028] 2. This invention considers both radial and tangential components when acquiring electromagnetic force data, and explicitly calculates the electromagnetic force phase difference between at least two sets of skewed pole segments at key frequencies and orders, generating a phase difference characterization quantity. This characterization quantity can accurately reflect the electromagnetic excitation interference effect caused by the multi-segment skewed pole structure, enabling the evaluation model to not only have high prediction accuracy, but also provide a clear physical basis for optimizing structural parameters such as skewed pole angle and number of segments.
[0029] 3. The NVH performance evaluation model adopted in this invention has a lightweight structure and low input dimension, which can be easily embedded into parameter optimization algorithms to achieve automatic design with NVH performance as the target. At the same time, the method is also applicable to vehicle system-level simulation platforms or digital twin systems, overcoming the shortcomings of traditional full-order models that are difficult to integrate due to high computational complexity, and expanding the application capabilities of NVH evaluation technology in intelligent design and online monitoring scenarios. Attached Figure Description
[0030] Figure 1 A flowchart illustrating a method for evaluating the NVH performance of a drive motor according to an embodiment of this application is shown.
[0031] Figure 2 A schematic diagram of the structure of a drive motor NVH performance evaluation device provided in an embodiment of this application is shown. Detailed Implementation
[0032] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.
[0033] In one embodiment, this application provides a method for evaluating the NVH performance of a drive motor, which includes the following steps:
[0034] S1. The time-domain electromagnetic force distribution data on the stator teeth of each skew pole section of the target motor is obtained through electromagnetic simulation. The time-domain electromagnetic force distribution data includes radial and tangential components.
[0035] S2, perform spatial domain equivalent synthesis of the electromagnetic forces of each skew pole segment to generate a global equivalent electromagnetic force time-domain signal characterizing the overall excitation of the motor;
[0036] S3, perform frequency domain-order joint analysis on the global equivalent electromagnetic force time domain signal, identify the target frequency point that has a significant impact on NVH performance, and extract the main spatial order and main time order corresponding to the target frequency point to obtain the synthetic excitation force amplitude corresponding to the main spatial order and main time order.
[0037] S4, based on the time-domain electromagnetic force distribution data described in S1, at the target frequency point and under the main spatial order and main time order, calculate the electromagnetic force phase difference between at least two sets of slant pole segments, and generate a phase difference characterization quantity.
[0038] S5, input the target frequency point, the amplitude of the synthetic excitation force and the phase difference characterization quantity into the pre-constructed NVH performance evaluation model, and output the corresponding NVH performance index.
[0039] This application employs spatial domain equivalent synthesis of the electromagnetic forces in each skewed pole segment, combined with frequency domain-order joint analysis, to extract the target frequency point, the amplitude of the synthesized excitation force corresponding to the principal order, and the phase difference characterization between the skewed pole segments as low-dimensional input features, avoiding direct solution of the high-degree-of-freedom full-order model. The NVH performance evaluation model constructed in this way can achieve millisecond-level predictions on ordinary computing devices, significantly shortening the single evaluation time and meeting the engineering needs of rapid parameterization research and multi-scheme comparison in the early stages of motor design. Furthermore, the core physical information affecting NVH performance is preserved during the order reduction process, including the vector characteristics of the electromagnetic force, the excitation amplitude of the dominant spatiotemporal order, and the phase relationship between the skewed pole segments. This significantly reduces computational resource consumption while maintaining prediction accuracy comparable to full-order multiphysics simulation, achieving an effective balance between efficiency and accuracy.
[0040] This application considers both radial and tangential components when acquiring electromagnetic force data, and explicitly calculates the electromagnetic force phase difference between at least two sets of skewed pole segments at key frequencies and orders, generating a phase difference characterization quantity. This characterization quantity can accurately reflect the electromagnetic excitation interference effect caused by the multi-segment skewed pole structure, enabling the evaluation model to not only have high prediction accuracy, but also provide a clear physical basis for optimizing structural parameters such as skewed pole angle and number of segments.
[0041] In a preferred embodiment of this application, the NVH performance indicators include acoustic power or equivalent radiated acoustic power.
[0042] Among them, acoustic power is an internationally recognized objective measure of noise source intensity, directly reflecting the ability of a motor to radiate noise energy into the surrounding environment; while equivalent radiated acoustic power (ERP) is an approximation of acoustic power obtained based on structural vibration inversion, and is widely used in motor NVH simulation and testing. Using acoustic power or equivalent radiated acoustic power as the output NVH performance index allows the evaluation results of this application to be directly used for engineering benchmarking, regulatory compliance judgment, and product acceptance, avoiding secondary conversion errors and interpretation ambiguities caused by using intermediate physical quantities such as vibration acceleration.
[0043] In a preferred embodiment of this application, the NVH performance evaluation model is obtained by training with multiphysics coupling simulation data or measured test data and is constructed using a machine learning model.
[0044] This preferred embodiment utilizes multiphysics coupled simulation or measured data to train the machine learning model offline. This allows the NVH performance evaluation model to predict NVH performance at extremely low computational cost (milliseconds) after training, avoiding the need for time-consuming full-order multiphysics simulations for each evaluation. Furthermore, because the training data comes from high-fidelity simulations or real tests, the NVH performance evaluation model maintains prediction accuracy comparable to full-order methods within key frequency bands, effectively solving the core problems of high computational complexity and high resource consumption in existing technologies.
[0045] Furthermore, the machine learning models described in this application include support vector machine (SVM) models, neural network models, or Gaussian process regression models. SVMs are suitable for scenarios with small samples and high-dimensional features, exhibiting good generalization ability. Neural network models train and predict quickly with medium-sized datasets, making them suitable for embedded deployments. Gaussian process regression models provide probabilistic outputs, suitable for reliability analysis where prediction confidence is required. By providing multiple optional model architectures, this implementation can be flexibly configured according to specific engineering resources, data scale, and accuracy requirements, enhancing the universality and practicality of the technical solution.
[0046] The NVH performance evaluation model adopted in this application has a lightweight structure and low input dimension, which can be easily embedded into parameter optimization algorithms to achieve automatic design with NVH performance as the goal. At the same time, the method is also applicable to whole vehicle system-level simulation platforms or digital twin systems, overcoming the shortcomings of traditional full-order models that are difficult to integrate due to high computational complexity, and expanding the application capabilities of NVH evaluation technology in intelligent design and online monitoring scenarios.
[0047] The specific training process of the NVH performance evaluation model includes the following steps:
[0048] A101, Constructing a training sample set: For multiple drive motor design schemes with different skew pole structure parameters, including the number of skew pole segments, skew pole angle, and axial segment position, perform the following operations respectively:
[0049] a. Obtain time-domain electromagnetic force distribution data on stator teeth of each skewed pole section through electromagnetic field-structure field coupling simulation or prototype testing.
[0050] b. Extract the target frequency point from the time-domain electromagnetic force distribution data according to the method described above in this application, i.e., steps S2 to S4 in claim 1. The amplitude of the composite excitation force under the corresponding principal order and the phase difference characterization between skew pole segments. .
[0051] c. Obtain the actual NVH performance indicators of the motor at the target frequency point through full-order acoustic simulation or actual measurement of sound power / equivalent radiated sound power (ERP). .
[0052] d. Combine the triples As input features, As labels, they form a training sample; repeat the above process to form a training dataset containing hundreds to thousands of samples.
[0053] A102, Select and configure a machine learning model: Choose one from support vector machine, neural network, or Gaussian process regression models as the basic architecture; set the model hyperparameters according to the data scale and accuracy requirements.
[0054] A103, Model Training and Validation: The training sample set is divided into a training set and a validation set. The selected machine learning model is trained using the training set, and the model parameters are optimized by minimizing the loss function between the predicted values and the actual NVH performance indicators. The model's generalization ability is evaluated using the validation set, and overfitting is prevented through early stopping, regularization, or cross-validation strategies. For example, the loss function is the mean squared error function.
[0055] A104, Model Deployment and Application: The trained NVH performance evaluation model is solidified into a callable function or module. In subsequent evaluations of new motor design schemes, only the input extracted from steps S2–S4 needs to be applied. This feature allows for the rapid output of predicted NVH performance metrics without the need to run full-order multiphysics simulations or physical tests again.
[0056] In a preferred embodiment of this application, the neural network model is a BP (Back Propagation) neural network model; the input layer of the BP neural network model includes three neurons, which are used to receive the target frequency point, the amplitude of the synthesized excitation force, and the phase difference representation quantity, respectively; the output layer of the BP neural network model includes one neuron, which is used to output NVH performance indicators; the BP neural network model adopts a nonlinear activation function and is trained by a numerical optimization algorithm.
[0057] In this preferred embodiment, the input dimension of the BP neural network model strictly matches the three core physical features extracted in this application: the target frequency point, the amplitude of the synthetic excitation force, and the phase difference representation, forming a lightweight network structure of 3 hidden layers and 1 layer. This structure has fewer parameters and faster training speed, effectively avoiding the overfitting risk caused by high-dimensional input, and is particularly suitable for motor NVH datasets with limited sample size, significantly improving the model's generalization performance in engineering practice. Simultaneously, due to the simple structure and fast inference speed of this BP neural network model, it can be seamlessly integrated into motor parameter optimization systems as a high-fidelity surrogate model, such as for tuning the slant pole angle and pole arc coefficient. Its single prediction time is typically in the millisecond range, enabling gradient-based or sampling-based optimization algorithms to complete hundreds to thousands of evaluation iterations within a reasonable time, significantly improving the automation level of NVH-oriented design.
[0058] Furthermore, the NVH performance indicators of drive motors, such as acoustic power or ERP, exhibit a strong nonlinear relationship with electromagnetic excitation parameters, especially showing sensitive and non-monotonic variations in the structural resonance region. Backpropagation (BP) neural networks, by introducing nonlinear activation functions, can effectively fit such complex mapping relationships, thereby achieving high-precision predictions within key frequency bands, outperforming simplified models such as linear regression. For example, the nonlinear activation functions include Sigmoid, Tanh, or ReLU.
[0059] Furthermore, mature numerical optimization algorithms (such as L-BFGS, Adam, or quasi-Newton methods) are used to update the network weights to ensure stable convergence during the training process. At the same time, the BP neural network architecture is mature, highly compatible, and easy to deploy in MATLAB, Python, or embedded platforms, meeting the application needs of multiple scenarios from R&D simulation to online evaluation.
[0060] In a preferred embodiment of this application, the frequency domain-order joint analysis is implemented using a two-dimensional Fourier transform, which is used to simultaneously analyze the harmonic characteristics of electromagnetic excitation in the time and space dimensions.
[0061] This preferred embodiment employs a two-dimensional Fourier transform to simultaneously expand the global equivalent electromagnetic force time-domain signal in both the time and spatial domains, generating a frequency-spatial order spectrum. This accurately identifies excitation components that simultaneously possess high amplitude, specific time frequency, and specific spatial order. These components are most likely to induce stator structure resonance, which is the main cause of NVH problems. This method significantly improves the accuracy of principal order extraction.
[0062] Furthermore, the discretized frequency-order grid provided by the two-dimensional Fourier transform provides an accurate frequency and spatial domain positioning basis for phase difference calculation, avoiding phase misjudgment caused by order aliasing or frequency shift, and ensuring the physical reliability of the phase difference representation.
[0063] In one embodiment, this application provides a drive motor NVH performance evaluation device, see [link to relevant documentation]. Figure 2 As shown, the evaluation device 10 includes an electromagnetic force data acquisition module 11, an excitation synthesis module 12, a principal order feature extraction module 13, a phase difference calculation module 14, and an evaluation model module 15.
[0064] The electromagnetic force data acquisition module 11 is used to perform the following: acquire time-domain electromagnetic force distribution data on the stator teeth of each slant pole section of the target motor through electromagnetic simulation, wherein the time-domain electromagnetic force distribution data includes radial and tangential components.
[0065] The excitation synthesis module 12 is used to perform: spatial domain equivalent synthesis of the electromagnetic force of each skew pole segment to generate a global equivalent electromagnetic force time domain signal characterizing the overall excitation of the motor.
[0066] The main order feature extraction module 13 is used to perform: frequency domain-order joint analysis on the global equivalent electromagnetic force time domain signal, identify the target frequency point that has a significant impact on NVH performance, and extract the main spatial order and main time order corresponding to the target frequency point to obtain the synthetic excitation force amplitude corresponding to the main spatial order and main time order.
[0067] The phase difference calculation module 14 is used to perform the following: based on the time-domain electromagnetic force distribution data acquired by the electromagnetic force data acquisition module, at the target frequency point and the main spatial order and the main time order, calculate the electromagnetic force phase difference between at least two sets of slant pole segments, and generate a phase difference characterization quantity.
[0068] The evaluation model module 15 is used to perform the following: inputting the target frequency point, the amplitude of the synthetic excitation force, and the phase difference characterization quantity into the pre-constructed NVH performance evaluation model, and outputting the corresponding NVH performance indicators.
[0069] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments, and various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention. Furthermore, the embodiments of the present invention and the features thereof can be combined with each other unless otherwise specified.
Claims
1. A method for evaluating the NVH performance of a drive motor, characterized in that, Includes the following steps: S1. The time-domain electromagnetic force distribution data on the stator teeth of each skew pole section of the target motor is obtained through electromagnetic simulation. The time-domain electromagnetic force distribution data includes radial and tangential components. S2, perform spatial domain equivalent synthesis of the electromagnetic forces of each skew pole segment to generate a global equivalent electromagnetic force time-domain signal characterizing the overall excitation of the motor; S3, perform frequency domain-order joint analysis on the global equivalent electromagnetic force time domain signal, identify the target frequency point that has a significant impact on NVH performance, and extract the main spatial order and main time order corresponding to the target frequency point to obtain the synthetic excitation force amplitude corresponding to the main spatial order and main time order. S4, based on the time-domain electromagnetic force distribution data described in S1, at the target frequency point and under the main spatial order and main time order, calculate the electromagnetic force phase difference between at least two sets of slant pole segments, and generate a phase difference characterization quantity. S5, input the target frequency point, the amplitude of the synthetic excitation force and the phase difference characterization quantity into the pre-constructed NVH performance evaluation model, and output the corresponding NVH performance index.
2. The method for evaluating the NVH performance of a drive motor according to claim 1, characterized in that: The NVH performance indicators include acoustic power or equivalent radiated acoustic power.
3. The method for evaluating the NVH performance of a drive motor according to claim 1, characterized in that: The NVH performance evaluation model is obtained by training with multi-physics field coupled simulation data or measured test data, and is constructed using a machine learning model. The machine learning model includes a support vector machine model, a neural network model, or a Gaussian process regression model.
4. The method for evaluating the NVH performance of a drive motor according to claim 3, characterized in that: The neural network model is a BP neural network model; The input layer of the BP neural network model includes three neurons, which are used to receive the target frequency point, the amplitude of the synthetic excitation force, and the phase difference representation, respectively. The output layer of the BP neural network model includes one neuron for outputting NVH performance metrics; The BP neural network model employs a nonlinear activation function and is trained using a numerical optimization algorithm.
5. The method for evaluating the NVH performance of a drive motor according to claim 1, characterized in that: The frequency domain-order joint analysis is implemented using a two-dimensional Fourier transform, which is used to simultaneously analyze the harmonic characteristics of electromagnetic excitation in the time and space dimensions.
6. A device for evaluating the NVH performance of a drive motor, characterized in that, include: The electromagnetic force data acquisition module is used to perform the following: acquiring time-domain electromagnetic force distribution data on the stator teeth of each skew pole section of the target motor through electromagnetic simulation, wherein the time-domain electromagnetic force distribution data includes radial and tangential components; The excitation synthesis module is used to perform: spatial domain equivalent synthesis of the electromagnetic force of each skewed pole segment, and generate a global equivalent electromagnetic force time-domain signal characterizing the overall excitation of the motor; The main order feature extraction module is used to perform: frequency domain-order joint analysis on the global equivalent electromagnetic force time domain signal, identify the target frequency point that has a significant impact on NVH performance, and extract the main spatial order and main time order corresponding to the target frequency point to obtain the synthetic excitation force amplitude corresponding to the main spatial order and main time order; The phase difference calculation module is used to perform the following: based on the time-domain electromagnetic force distribution data acquired by the electromagnetic force data acquisition module, at the target frequency point and the main spatial order and the main time order, calculate the electromagnetic force phase difference between at least two sets of slant pole segments, and generate a phase difference characterization quantity. The evaluation model module is used to perform the following: inputting the target frequency point, the amplitude of the synthetic excitation force, and the phase difference characterization quantity into the pre-built NVH performance evaluation model, and outputting the corresponding NVH performance indicators.