A permanent magnet synchronous motor multi-type demagnetization fault diagnosis method and system

Abstract

The application provides a permanent magnet synchronous motor multi-type demagnetization fault diagnosis method and system, relates to the motor fault diagnosis field, and solves the technical problems that the existing technology does not depend on an accurate model, but the extracted features often have insufficient distinction, and it is difficult to accurately determine the fault type and locate the fault magnetic pole. The method comprises the following steps: obtaining the air gap magnetic density signal of the permanent magnet synchronous motor under different demagnetization faults; obtaining the magnetic density difference signal based on the air gap magnetic density signal and the reference air gap magnetic density signal; dividing the magnetic density difference signal waveform into a plurality of waveform regions, calculating the area of the waveform region corresponding to each magnetic pole, obtaining the air gap magnetic density waveform area difference, and calculating the fault characteristic quantity of each magnetic pole to obtain a multi-dimensional fault feature vector; constructing a fault classification model based on the multi-dimensional feature vector and the corresponding fault type; and constructing a demagnetization degree prediction model based on the air gap magnetic density signal and the corresponding fault degree. The application is used in the motor fault diagnosis process.

Description

Technical Field

[0001] This application relates to the field of motor fault diagnosis, and in particular to a method and system for diagnosing multiple types of demagnetization faults in permanent magnet synchronous motors. Background Technology

[0002] Permanent magnet synchronous motors (PMSMs) are widely used in important fields such as electric vehicles and industrial automation due to their high efficiency and stability. Their performance hinges on the magnetic field of the permanent magnets; however, permanent magnets are prone to demagnetization under conditions such as high temperature and oxidation, leading to performance degradation, reduced efficiency, and even safety hazards. Therefore, conducting research on early demagnetization fault diagnosis is crucial.

[0003] Existing diagnostic methods based on fault signal feature extraction, although not dependent on precise models, often lack sufficient discriminative power in the extracted features. For example, signal features such as torque and vibration have weak correlation with demagnetization faults, making it difficult to accurately determine the fault type and locate the faulty magnetic pole. Summary of the Invention

[0004] This application provides a method for diagnosing multiple types of demagnetization faults in permanent magnet synchronous motors, solving the technical problem that while existing technologies do not rely on precise models, the extracted features often lack sufficient discriminative power, making it difficult to accurately determine the fault type and locate the faulty magnetic pole. This application also provides a system for diagnosing multiple types of demagnetization faults in permanent magnet synchronous motors.

[0005] To achieve the above objectives, this application adopts the following technical solution:

[0006] Firstly, a method for diagnosing multiple types of demagnetization faults in permanent magnet synchronous motors is provided, including:

[0007] Acquire air gap magnetic flux density signals of permanent magnet synchronous motors under different demagnetization faults; demagnetization faults include fault type and fault severity;

[0008] Based on the air gap magnetic flux density signal of the demagnetization fault and the reference air gap magnetic flux density signal, the magnetic flux density difference signal is obtained; wherein, the reference air gap magnetic flux density signal is the air gap magnetic flux density signal of the permanent magnet synchronous motor in a healthy state.

[0009] The magnetic flux density difference signal waveform is divided into several waveform regions. The area of ​​the waveform region corresponding to each magnetic pole is calculated to obtain the air gap magnetic flux density waveform area difference. Based on the air gap magnetic flux density waveform area difference, the fault characteristic quantity of each magnetic pole is calculated to obtain a multidimensional fault characteristic vector.

[0010] A fault classification model is constructed based on multidimensional feature vectors and corresponding fault types; the fault classification model is based on the PSO-SVM fault classification model.

[0011] Furthermore, a demagnetization degree prediction model is constructed based on the air gap magnetic flux density signal and the corresponding fault degree; the demagnetization degree prediction model is constructed based on the CNN-Bilstm-Attention model.

[0012] Based on the above technical solution, a dual-model collaborative architecture of "PSO-SVM + CNN-BiLSTM-Attention" is adopted to accurately address the two core requirements of demagnetization fault type identification and severity quantification. Specifically, for the fault classification task, since the number of fault samples in actual engineering is limited and the feature dimensions are high, traditional classifiers are prone to overfitting or rely on manual parameter tuning. Therefore, SVM is selected, as it has excellent generalization ability in small sample and high-dimensional spaces. Simultaneously, the PSO algorithm is introduced to automatically optimize the penalty coefficient and kernel parameters of the SVM, avoiding the high computational cost and local optimum traps of grid search, significantly improving the identification accuracy and robustness of multiple types of demagnetization faults, such as unipolar, multipolar, and symmetric / asymmetric demagnetization. On the other hand, for the continuous regression problem of predicting the degree of demagnetization, a single model struggles to simultaneously capture local abrupt changes, long-range temporal dependencies, and critical fault periods in the magnetic flux density signal. Therefore, a CNN-BiLSTM-Attention model is constructed: the CNN layer automatically extracts local spatial features of the magnetic flux density waveform, such as harmonic distortion and peak shift; the BiLSTM layer models forward and reverse temporal dynamics to capture the demagnetization evolution pattern; and the attention mechanism dynamically weights key time steps, strengthening the focus on demagnetization-sensitive sections, thereby achieving high-precision, end-to-end prediction of the severity of demagnetization. This combination of technologies not only avoids tedious manual feature engineering but also solves key technical challenges such as "difficulty in distinguishing multiple types of faults," "discontinuous degree assessment," and "low efficiency of online diagnosis" through the complementarity of feature-driven and data-driven approaches, combining high precision, strong adaptability, and engineering practicality.

[0013] In conjunction with the first aspect above, in one possible implementation, acquiring the air gap magnetic flux density signal of the permanent magnet synchronous motor under different demagnetization faults includes:

[0014] Establish a finite element model of the permanent magnet synchronous motor;

[0015] Based on the demagnetization principle of permanent magnets, the coercivity of the permanent magnets in the finite element model of the permanent magnet synchronous motor is changed to simulate the finite element model of different demagnetization faults, and the air gap magnetic flux density signal of different demagnetization faults is extracted.

[0016] In conjunction with the first aspect above, in one possible implementation, acquiring the magnetic density difference signal includes:

[0017] Align the rotor position of the air gap magnetic flux density signal to be diagnosed with that of the reference air gap magnetic flux density signal, and subtract the amplitude of the corresponding rotor position to obtain the magnetic flux density difference signal.

[0018] In conjunction with the first aspect above, in one possible implementation, the method for dividing the magnetic density difference signal waveform into several waveform regions includes:

[0019] The sliding window method is used to divide the magnetic density difference signal within one mechanical cycle into several continuous and non-overlapping time windows, resulting in several waveform regions; the number of time windows is equal to the number of magnetic poles of the permanent magnet synchronous motor.

[0020] In conjunction with the first aspect above, in one possible implementation, the calculation of fault characteristic quantities for each magnetic pole... ,include:

[0021] ;

[0022] in, Let be the effective length of the permanent magnet motor core, A be the area of ​​the cross-section of the permanent magnet core, and i be the pole number. Let t0 be the area difference of the air gap magnetic flux density waveform under the i-th magnetic pole, t0 be the starting position of the time window, and t1 be the ending position of the time window.

[0023] In conjunction with the first aspect above, in one possible implementation, the area difference of the air gap magnetic flux density waveform under the i-th magnetic pole is... ,include:

[0024] ;

[0025] Where j is the sampling point number, For the first Within the time window corresponding to the first magnetic pole The magnetic flux density difference at each sampling point For the first The angle interval corresponding to the nth sampling point, N is the nth sampling point. The total number of sampling points within the time window corresponding to each magnetic pole.

[0026] In conjunction with the first aspect above, in one possible implementation, the fault classification model is constructed based on the PSO-SVM fault classification model, including:

[0027] Extract multidimensional feature vectors and corresponding fault type labels from historical data for several fault states to construct a demagnetization fault sample library;

[0028] The multidimensional feature vectors in the demagnetization sample library are used as input data for the PSO-SVM fault classification model, and the fault type labels are used as output data for the PSO-SVM fault classification model. The PSO-SVM fault classification model is trained by using the input and output data to obtain the fault classification model.

[0029] In conjunction with the first aspect above, in one possible implementation, the CNN-BiLSTM-Attention model includes a convolutional neural network layer, a bidirectional long short-term memory network layer, and an attention mechanism layer;

[0030] Convolutional neural network layers are used to extract local features from air gap magnetic flux density signal data;

[0031] Bidirectional long short-term memory network layers are used to learn the temporal dependencies of features after processing by convolutional neural network layers;

[0032] The attention mechanism layer is used to assign differentiated weights to the output sequence of the bidirectional long short-term memory network layer.

[0033] In conjunction with the first aspect above, in one possible implementation, the demagnetization degree prediction model is constructed based on the CNN-Bilstm-Attention model, including:

[0034] The air gap magnetic flux density signal data of permanent magnet synchronous motor under different demagnetization degrees were obtained from historical data, and the corresponding true values ​​of demagnetization degree were marked.

[0035] The air gap magnetic flux density signal data is used as the input data of the CNN-BiLSTM-Attention model, and the true value of the demagnetization degree is used as the output data of the CNN-BiLSTM-Attention model. The CNN-BiLSTM-Attention model is trained by using the input data and the output data to obtain the demagnetization degree prediction model.

[0036] Secondly, a multi-type demagnetization fault diagnosis device for permanent magnet synchronous motors is provided, comprising: a communication unit and a processing unit; the communication unit is used to acquire air gap magnetic flux density signals of the permanent magnet synchronous motor under different demagnetization faults; the processing unit is used to acquire a magnetic flux density difference signal based on the air gap magnetic flux density signal of the demagnetization fault and a reference air gap magnetic flux density signal; divide the waveform of the magnetic flux density difference signal into several waveform regions, calculate the area of ​​the waveform region corresponding to each magnetic pole, and obtain the air gap magnetic flux density waveform area difference; calculate the fault characteristic quantity of each magnetic pole based on the air gap magnetic flux density waveform area difference, and obtain a multi-dimensional fault feature vector; and construct a fault classification model based on the multi-dimensional feature vector and the corresponding fault type; and construct a demagnetization degree prediction model based on the air gap magnetic flux density signal and the corresponding fault degree.

[0037] Thirdly, this application provides a multi-type demagnetization fault diagnosis device for permanent magnet synchronous motors, comprising: a processor and a storage medium; the storage medium includes instructions, and the processor is used to execute the instructions to implement the method described in the first aspect and any possible implementation thereof. This multi-type demagnetization fault diagnosis device for permanent magnet synchronous motors can be an electronic device or a chip within an electronic device.

[0038] Fourthly, this application provides a multi-type demagnetization fault diagnosis system for permanent magnet synchronous motors, comprising: an acquisition module, a feature extraction module, and a fault model construction module; wherein, the acquisition module is used to acquire air gap magnetic flux density signals of the permanent magnet synchronous motor under different demagnetization faults; the feature extraction module is used to acquire magnetic flux density difference signals based on the air gap magnetic flux density signals under demagnetization faults and reference air gap magnetic flux density signals; divide the waveform of the magnetic flux density difference signal into several waveform regions, calculate the area of ​​the waveform region corresponding to each magnetic pole, and obtain the air gap magnetic flux density waveform area difference; calculate the fault feature quantity of each magnetic pole based on the air gap magnetic flux density waveform area difference, and obtain a multi-dimensional fault feature vector; the fault model construction module constructs a fault classification model based on the multi-dimensional feature vector and the corresponding fault type; and constructs a demagnetization degree prediction model based on the air gap magnetic flux density signal and the corresponding fault degree.

[0039] Fifthly, this application provides a computer-readable storage medium storing instructions that, when executed on a permanent magnet synchronous motor multi-type demagnetization fault diagnosis device, cause the permanent magnet synchronous motor multi-type demagnetization fault diagnosis device to perform the method described in the first aspect and any possible implementation thereof.

[0040] Sixthly, this application provides a computer program product containing instructions that, when the computer program product is run on a permanent magnet synchronous motor multi-type demagnetization fault diagnosis device, cause the permanent magnet synchronous motor multi-type demagnetization fault diagnosis device to perform the method described in the first aspect and any possible implementation of the first aspect.

[0041] This application provides a method and system for diagnosing multiple types of demagnetization faults in permanent magnet synchronous motors. By introducing the magnetic flux density difference signal and combining it with the sliding window method to extract the area difference of the air gap magnetic flux density waveform corresponding to each magnetic pole as a key feature, it effectively highlights the local magnetic field distortion caused by demagnetization, significantly reduces the feature dimension, avoids high-dimensional computational redundancy, and greatly reduces the computational burden of subsequent classification algorithms while ensuring diagnostic accuracy, thus improving the model response speed and providing strong support for online real-time diagnosis. Secondly, this scheme uses PSO-SVM to construct a fault classification model, making full use of the strong classification ability of SVM under small samples and high-dimensional features and the global optimization characteristics of the PSO algorithm, achieving high accuracy in identifying various types of demagnetization faults (such as uniform demagnetization, local demagnetization, asymmetric demagnetization, etc.), demonstrating excellent robustness and classification performance. At the same time, a demagnetization degree prediction model based on CNN-BiLSTM-Attention is further constructed. Utilizing its powerful end-to-end nonlinear modeling capability, it automatically extracts the local spatial features and long-term temporal dependencies of the air gap magnetic flux density signal, and focuses on key fault periods through an attention mechanism, achieving continuous and accurate quantitative prediction of the demagnetization degree. The overall solution can not only accurately identify the fault type, but also assess the severity of the fault. It has both multi-functionality and engineering universality, and is applicable to intelligent diagnosis of demagnetization faults of permanent magnet synchronous motors with different pole numbers and operating conditions. It has good application prospects and promotion value.

[0042] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0043] Figure 1 A system architecture diagram of a multi-type demagnetization fault diagnosis system for permanent magnet synchronous motors provided in this application embodiment;

[0044] Figure 2 A flowchart illustrating a method for diagnosing multiple types of demagnetization faults in a permanent magnet synchronous motor, provided in an embodiment of this application;

[0045] Figure 3A schematic diagram illustrating the construction process of a fault severity assessment model provided in this application embodiment;

[0046] Figure 4 A schematic diagram of the air gap magnetic flux density waveform area difference for each magnetic pole of the permanent magnet synchronous motor provided in this application under demagnetization fault type 1;

[0047] Figure 5 A schematic diagram of the air gap magnetic flux density waveform area difference for each magnetic pole of the permanent magnet synchronous motor provided in this application under demagnetization fault type 2;

[0048] Figure 6 A schematic diagram of the air gap magnetic flux density waveform area difference for each magnetic pole of the permanent magnet synchronous motor provided in this application under demagnetization fault type 3;

[0049] Figure 7 A schematic diagram of the air gap magnetic flux density waveform area difference for each magnetic pole of the permanent magnet synchronous motor provided in this application under demagnetization fault type 4;

[0050] Figure 8 A schematic diagram of the air gap magnetic flux density waveform area difference for each magnetic pole of the permanent magnet synchronous motor provided in this application under demagnetization fault type 5;

[0051] Figure 9 A schematic diagram of the air gap magnetic flux density waveform area difference for each magnetic pole of the permanent magnet synchronous motor provided in this application under demagnetization fault type 6;

[0052] Figure 10 A schematic diagram of the air gap magnetic flux density waveform area difference for each magnetic pole of the permanent magnet synchronous motor provided in this application under demagnetization fault type 7;

[0053] Figure 11 A schematic diagram of the air gap magnetic flux density waveform area difference for each magnetic pole of the permanent magnet synchronous motor provided in this application under demagnetization fault type 8;

[0054] Figure 12 A schematic diagram of the air gap magnetic flux density waveform area difference for each magnetic pole of the permanent magnet synchronous motor provided in this application under demagnetization fault type 9;

[0055] Figure 13 A schematic diagram of the air gap magnetic flux density waveform area difference for each magnetic pole of the permanent magnet synchronous motor provided in this application under demagnetization fault type 10;

[0056] Figure 14 A schematic diagram of the air gap magnetic flux density waveform area difference for each magnetic pole of the permanent magnet synchronous motor provided in this application under demagnetization fault type 11;

[0057] Figure 15This is a schematic diagram of the structure of a demagnetization fault diagnosis device provided in an embodiment of this application;

[0058] Figure 16 This is a schematic diagram of the hardware structure of a demagnetization fault diagnosis device provided in an embodiment of this application. Detailed Implementation

[0059] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] The method for diagnosing multiple types of demagnetization faults in a permanent magnet synchronous motor provided in this application embodiment can be applied to a system for diagnosing multiple types of demagnetization faults in a permanent magnet synchronous motor, such as... Figure 1 As shown, the communication system includes: an acquisition module 100, a feature extraction module 200, and a fault model construction module 300;

[0061] The acquisition module 100 is used to acquire the air gap magnetic flux density signal of the permanent magnet synchronous motor under different demagnetization faults.

[0062] The feature extraction module 200 is used to obtain the magnetic density difference signal based on the air gap magnetic density signal of the demagnetization fault and the reference air gap magnetic density signal; divide the waveform of the magnetic density difference signal into several waveform regions, calculate the area of ​​the waveform region corresponding to each magnetic pole, and obtain the air gap magnetic density waveform area difference; and calculate the fault feature quantity of each magnetic pole based on the air gap magnetic density waveform area difference to obtain a multi-dimensional fault feature vector.

[0063] Fault model construction module 300: Based on multi-dimensional feature vectors and corresponding fault types, a fault classification model is constructed; and based on air gap magnetic flux density signals and corresponding fault degrees, a demagnetization degree prediction model is constructed.

[0064] To address the technical problem that existing technologies, while not relying on precise models, often suffer from insufficient feature discrimination in their extracted features, making it difficult to accurately determine fault types and locate faulty magnetic poles, this application provides a method for diagnosing multiple types of demagnetization faults in permanent magnet synchronous motors. This method includes: acquiring air gap magnetic flux density signals of the permanent magnet synchronous motor under different demagnetization faults; acquiring a magnetic flux density difference signal based on the air gap magnetic flux density signal under demagnetization faults and a reference air gap magnetic flux density signal; wherein the reference air gap magnetic flux density signal is the air gap magnetic flux density signal of the permanent magnet synchronous motor in a healthy state; dividing the waveform of the magnetic flux density difference signal into several waveform regions, and calculating the waveform corresponding to each magnetic pole. The area of ​​the shaped region is used to obtain the area difference of the air gap magnetic flux density waveform. Based on the area difference of the air gap magnetic flux density waveform, the fault feature quantity of each magnetic pole is calculated to obtain a multi-dimensional fault feature vector. Based on the multi-dimensional feature vector and the corresponding fault type, a fault classification model is constructed. Furthermore, based on the air gap magnetic flux density signal and the corresponding fault degree, a demagnetization degree prediction model is constructed. Based on this, by constructing the magnetic flux density difference signal and dividing the corresponding region of the magnetic pole, the area difference of the air gap magnetic flux density waveform is extracted as the core feature, effectively reflecting the magnetic field anomaly caused by local demagnetization. The feature has clear physical meaning and low dimensionality, significantly improving computational efficiency and supporting real-time online diagnosis. A PSO-optimized SVM model is used for fault classification, which can accurately identify multiple demagnetization types even with limited samples, demonstrating strong adaptability and stability. Simultaneously, a CNN-BiLSTM-Attention network is introduced to learn the original magnetic flux density signal end-to-end, integrating spatial feature extraction from convolutional layers, bidirectional temporal modeling, and attention weight allocation to achieve fine-grained quantitative prediction of the demagnetization degree. The overall method has both fault type identification and severity assessment capabilities, does not rely on complex manual feature engineering, is applicable to multi-condition and multi-pole motors, provides comprehensive and high-precision diagnosis, and has good versatility and practical application potential.

[0065] like Figure 2 As shown in the embodiment of this application, a method for diagnosing multiple types of demagnetization faults in a permanent magnet synchronous motor includes:

[0066] S201. Obtain the air gap magnetic flux density signal of the permanent magnet synchronous motor under different demagnetization faults.

[0067] Demagnetization faults include fault type and fault severity.

[0068] S202. Based on the air gap magnetic flux density signal of the demagnetization fault and the reference air gap magnetic flux density signal, obtain the magnetic flux density difference signal.

[0069] Among them, the reference air gap magnetic flux density signal is the air gap magnetic flux density signal of a permanent magnet synchronous motor in a healthy state.

[0070] S203. Divide the magnetic density difference signal waveform into several waveform regions, calculate the area of ​​the waveform region corresponding to each magnetic pole, and obtain the air gap magnetic density waveform area difference.

[0071] S204. Based on the area difference of the air gap magnetic flux density waveform, calculate the fault characteristic quantities of each magnetic pole to obtain a multidimensional fault characteristic vector.

[0072] It should be noted that the first Magnetic density difference value of magnetic density difference signal under magnetic pole The difference in waveform area between the air gap magnetic flux density and the magnetic pole. It is directly proportional, and the specific derivation process is as follows:

[0073] Let the first The total number of sampling points within the time window corresponding to each magnetic pole is The outer diameter arc length of the rotor is The effective length of the permanent magnet motor core is According to the basic principles of electromagnetism, the first The magnetic flux at the poles can be expressed as:

[0074] ;

[0075] in, For the first The area of ​​the lower air gap magnetic flux density waveform, i.e., the integral of the magnetic flux density amplitude with respect to the angle. This represents the actual magnetic flux.

[0076] From the change in magnetic flux, we can obtain: ;in, The reference magnetic flux under healthy conditions;

[0077] Further, introduce magnetic flux difference per unit area: Where A is the area of ​​the cross-section of the permanent magnet core;

[0078] in, This is the cross-sectional area of ​​the air gap;

[0079] Therefore, the first Magnetic flux density difference under magnetic poles: .

[0080] S205. Based on multidimensional feature vectors and corresponding fault types, a fault classification model is constructed; wherein, the fault classification model is constructed based on the PSO-SVM fault classification model.

[0081] Furthermore, a demagnetization degree prediction model is constructed based on the air gap magnetic flux density signal and the corresponding fault degree; the demagnetization degree prediction model is constructed based on the CNN-Bilstm-Attention model.

[0082] In one possible implementation of this application embodiment, the above-mentioned S201 can be specifically described as follows:

[0083] Build a finite element simulation model of a permanent magnet synchronous motor in Maxwell;

[0084] Based on the demagnetization principle of permanent magnets, the coercivity of the permanent magnets in the finite element model of the permanent magnet synchronous motor is changed to simulate the finite element model of different demagnetization faults, and the air gap magnetic flux density signal of different demagnetization faults is extracted.

[0085] The air gap magnetic flux density signal can be obtained in the Maxwell software. The signal is a periodic signal and is uniformly distributed.

[0086] In one possible implementation of this application embodiment, the above-mentioned S202 can be specifically implemented by the following S301, S302 and S303, which are described in detail below:

[0087] Align the rotor position of the air gap magnetic flux density signal to be diagnosed with that of the reference air gap magnetic flux density signal, and subtract the amplitude of the corresponding rotor position to obtain the magnetic flux density difference signal.

[0088] It should be noted that in the operation or simulation of a permanent magnet synchronous motor, the air gap magnetic flux density signal does not exist in isolation. It is measured or calculated during the dynamic process of rotor rotation. Therefore, each data point in the signal corresponds to a specific instantaneous spatial position of the rotor.

[0089] In one possible implementation of this application embodiment, the above-mentioned S203 can be specifically implemented by the following S301 and S302, which are described in detail below:

[0090] S301. Using the sliding window method, the magnetic density difference signal within one mechanical cycle is divided into several continuous and non-overlapping time windows to obtain several waveform regions.

[0091] The number of time windows is equal to the number of magnetic poles of the permanent magnet synchronous motor.

[0092] It should be noted that one mechanical cycle refers to one rotation of the motor rotor.

[0093] S302, the area difference of the air gap magnetic flux density waveform under the i-th magnetic pole ,include:

[0094] ;

[0095] Where j is the sampling point number, For the first Within the time window corresponding to the first magnetic pole The magnetic flux density difference at each sampling point For the first The angle interval corresponding to each sampling point.

[0096] It should be noted that this formula reflects the "cumulative error" of the magnetic field deviating from the ideal value under this magnetic pole, and is regarded as a discrete approximation of "area difference", that is... The larger the value, the more severe the magnetic field distortion in the region corresponding to the i-th magnetic pole, which may indicate that the magnetic pole has local demagnetization or assembly problems.

[0097] For example, a 4-pole permanent magnet synchronous motor → total number of poles = There are four N's ​​and four S's arranged alternately.

[0098] 800 sampling points were collected within one mechanical cycle, evenly distributed at 0. o ~360° mechanical angle;

[0099] The 800 sampling points were divided into 8 time windows (one window for each magnetic pole).

[0100] Each time window contains: Each sampling point corresponds to a 45° mechanical angle for each time window.

[0101] Taking the third time window (i=3) as an example, assuming it includes sampling points j=1 to 100, then The air gap magnetic flux density waveform area difference for the remaining time windows is calculated in this manner.

[0102] Based on the above technical solutions, this approach employs a sliding window division and magnetic flux density waveform area difference calculation technique to achieve high-precision, localized, and low-dimensional quantitative diagnosis of demagnetization faults in permanent magnet synchronous motors. Traditional fault diagnosis methods often rely on overall signal statistical characteristics, such as root mean square (RMS) and total harmonic distortion (THD), making it difficult to distinguish local magnetic pole anomalies, especially in multi-pole motors where fault location can be ambiguous. This method divides the magnetic flux density difference signal of one mechanical cycle (e.g., one rotor revolution) into several continuous and non-overlapping time windows based on the number of magnetic poles. Each window strictly corresponds to the spatial location of a physical magnetic pole, thus establishing a one-to-one mapping between fault characteristics and specific magnetic poles. Furthermore, by calculating the weighted sum of the absolute values ​​of the magnetic flux density difference within each window—the discretized area difference—the cumulative distortion degree of the magnetic field deviation from a healthy state in that pole region is effectively quantified. This indicator not only has clear physical meaning and can sensitively reflect typical faults such as local demagnetization, magnet misalignment, or assembly defects, but also avoids the problem of positive and negative deviations canceling each other out by taking absolute values, significantly improving the fault detection rate. More importantly, this method compresses the original high-dimensional time-series signal into a low-dimensional feature vector of the same length as the magnetic pole number, such as 8-dimensional, which significantly reduces the input dimensionality and computational complexity of the subsequent classification model, laying the foundation for efficient online diagnosis. In summary, this technology solves three key problems: "difficulty in fault location," "high feature redundancy," and "insensitivity to local demagnetization." It ensures diagnostic accuracy while also considering real-time performance and interpretability, demonstrating outstanding engineering application value.

[0103] In one possible implementation of this application embodiment, the above-mentioned S204 can be specifically described as follows:

[0104] Calculate the fault characteristic quantities of each magnetic pole. ,include:

[0105] ;

[0106] Where i is the magnetic pole number, t0 is the starting position of the time window, and t1 is the ending position of the time window.

[0107] In one possible implementation of this application embodiment, the above-mentioned S205 can be specifically described as follows:

[0108] The process of constructing the fault classification model is as follows:

[0109] Extract multidimensional feature vectors and corresponding fault type labels from historical data for several fault states to construct a demagnetization fault sample library;

[0110] The multidimensional feature vectors in the demagnetization sample library are used as input data for the PSO-SVM fault classification model, and the fault type labels are used as output data for the PSO-SVM fault classification model. The PSO-SVM fault classification model is trained by using the input and output data to obtain the fault classification model.

[0111] It should be noted that SVM (Support Vector Machine) excels at small-sample, high-dimensional classification, making it suitable for fault diagnosis scenarios.

[0112] Particle Swarm Optimization (PSO) is used to automatically search for optimal SVM parameters, such as penalty coefficients and kernel function parameters. The specific search process is as follows:

[0113] First, the penalty coefficient and kernel function parameters of the SVM are encoded as particle positions, and the classification accuracy is used as the fitness function to evaluate the performance of each particle. A group of particles is initialized using PSO, and their velocities and positions are iteratively updated in the solution space, continuously approximating the optimal parameter combination by combining individual historical best and global best information. In each iteration, the SVM is trained using the current parameters, and the fitness is calculated on the validation set until the convergence condition is met. Finally, the optimal parameters (penalty coefficient and kernel function parameters) are output, and the final fault classification model is constructed based on these parameters. This process effectively avoids the blindness of manual parameter tuning, significantly improves the accuracy and robustness of SVM in demagnetization fault classification, and realizes automated modeling from feature input to high-precision diagnosis, providing reliable technical support for intelligent fault diagnosis.

[0114] Based on the above technical solutions, this approach introduces fault feature quantities based on the integral of the area difference of magnetic flux density waveform, and combines it with PSO-SVM to construct a fault classification model. This aims to address key technical bottlenecks in permanent magnet synchronous motor (PMSM) demagnetization fault diagnosis, such as weak physical meaning of features, reliance on experience for parameter tuning, and difficulty in distinguishing between multiple fault types. First, by integrating the area difference of magnetic flux density over a time window and normalizing it using motor geometric parameters (core length and air gap cross-sectional area), the feature quantity acquires a clear electromagnetic physical meaning—directly reflecting the magnetic flux loss per unit area—thus more accurately characterizing the severity and spatial distribution of demagnetization at different magnetic poles. Second, considering the practical constraints of limited fault samples, numerous categories, and high feature dimensionality, SVM is chosen as the classifier due to its advantage in minimizing structural risk in small samples and high-dimensional spaces, effectively avoiding overfitting. However, SVM performance is highly dependent on the choice of penalty coefficient and kernel parameters, and traditional grid search is inefficient and prone to getting trapped in local optima. Therefore, a particle swarm optimization (PSO) algorithm is introduced, transforming parameter optimization into an intelligent search problem. Using classification accuracy as the fitness function, the algorithm automatically approximates the globally optimal parameter combination through swarm iteration. This strategy not only significantly improves the model's accuracy and robustness in identifying various demagnetization types, such as unipolar demagnetization, asymmetric demagnetization, and multipolar demagnetization, but also automates and intelligentizes the modeling process, greatly reducing reliance on human experience. The overall approach balances physical interpretability, computational efficiency, and high diagnostic accuracy, providing a reliable and practical technical path for online intelligent diagnosis of motor demagnetization faults.

[0115] Please see Figure 3 The process of constructing the fault severity assessment model is as follows:

[0116] The air gap magnetic flux density signal data of permanent magnet synchronous motor under different demagnetization degrees were obtained from historical data, and the corresponding true values ​​of demagnetization degree were marked.

[0117] The air gap magnetic flux density signal data is used as the input data of the CNN-BiLSTM-Attention model, and the true value of the demagnetization degree is used as the output data of the CNN-BiLSTM-Attention model. The CNN-BiLSTM-Attention model is trained by using the input data and the output data to obtain the demagnetization degree prediction model.

[0118] The CNN-BiLSTM-Attention model includes a convolutional neural network layer, a bidirectional long short-term memory network layer, and an attention mechanism layer.

[0119] A convolutional neural network layer is used to extract local features from the air gap magnetic flux density signal data and aggregate information and reduce feature dimensionality through pooling operations.

[0120] Bidirectional long short-term memory network layers are used to learn the temporal dependencies of features after processing by convolutional neural network layers;

[0121] An attention mechanism layer is used to assign differentiated weights to the output sequence of the bidirectional long short-term memory network layer to enhance attention to key abrupt changes in the signal.

[0122] Based on the above technical solutions, this approach employs a CNN-BiLSTM-Attention deep learning model to predict the demagnetization degree of a permanent magnet synchronous motor (PMSM). This aims to address the challenges faced by traditional methods in quantifying continuous demagnetization, such as weak feature representation, insufficient temporal modeling, and the tendency for key fault information to be obscured. Air gap magnetic flux density signals exhibit strong nonlinearity, high noise, and complex spatiotemporal coupling characteristics, making it difficult for a single model to simultaneously capture both local abrupt changes and global evolutionary patterns. Therefore, this architecture integrates the advantages of three neural networks: a Convolutional Neural Network (CNN) automatically extracts local spatial features from the magnetic flux density waveform, such as harmonic distortion, peak shift, and magnetic flux density dips, through sliding convolutional kernels, and utilizes pooling operations to suppress noise and reduce dimensionality; a Bidirectional Long Short-Term Memory (BiLSTM) network further models the feature sequence processed by the CNN, simultaneously learning forward and backward temporal dependencies to accurately depict the dynamic evolution of demagnetization from mild to severe; and an attention mechanism dynamically assigns weights to the temporal features output by the BiLSTM, automatically focusing on key time segments sensitive to demagnetization, such as magnetic pole switching points or anomalous abrupt change regions, effectively enhancing the model's ability to perceive subtle demagnetization changes. This end-to-end model eliminates the need for manually designed features, avoids reliance on experience, and directly maps the original magnetic flux density signal into continuous demagnetization degree values, such as demagnetization percentage, enabling high-precision, fine-grained assessment of demagnetization status. Compared to traditional regression methods or shallow models, this technology significantly improves prediction accuracy, robustness, and generalization ability, providing strong data-driven support for motor health monitoring and remaining life prediction.

[0123] Based on the technical solution of the present invention, the present invention Figures 4-14 The data presents the air gap magnetic flux density waveform area difference for each of the eight poles (out of a total) of a permanent magnet synchronous motor under 11 different demagnetization fault types. The distribution of the magnets is shown in the figure. Each sub-figure corresponds to a typical demagnetization fault mode. The horizontal axis represents the 1st to 8th permanent magnets, arranged in spatial order, and the vertical axis represents the corresponding area difference (unit: T·s), reflecting the degree of magnetic field distortion in each magnetic pole region.

[0124] in, Figure 4 Fault type 1: A significant negative area difference appears in the first magnetic pole, indicating that it has undergone severe demagnetization; Figure 5 Fault type 2: The second magnetic pole suddenly increases positively, which may be due to local overmagnetism or asymmetrical demagnetization; Figure 6Fault type 3: The third magnetic pole is significantly negatively biased, indicating localized demagnetization; Figure 7 Fault type 4: The fourth magnetic pole is abnormally positive, while the others are basically normal; Figure 8 Fault type 5: The third magnetic pole is prominently negative, while the others show slight fluctuations; Figure 9 Fault type 6: The 6th magnetic pole is positively enhanced, showing a unipolar change; Figure 10 Fault type 7: The 7th magnetic pole is significantly negative, indicating local demagnetization; Figure 11 Fault type 8: The 8th magnetic pole rises sharply in the positive direction, possibly due to heat or stress affecting the edge magnetic poles; Figure 12 Fault type 9: The first and third magnetic poles show positive and negative abnormalities respectively, exhibiting symmetrical or alternating demagnetization characteristics; Figure 13 For fault type 10: the second magnetic pole shows a significant positive area difference, the fifth magnetic pole shows a significant negative area difference, and the remaining magnetic poles show little change, indicating that there is asymmetric local demagnetization or a situation where some magnetic poles are overmagnetized / demagnetized simultaneously, which has strong local and directional characteristics. Figure 14 Fault type 11: Odd-numbered magnetic poles (1, 3, 5, 7) all show negative area differences, while even-numbered magnetic poles (2, 4, 6, 8) show positive area differences, forming an alternating, periodic distribution of magnetic field anomalies, which may correspond to overall asymmetric demagnetization or systematic deviations caused by batch differences in magnetic pole materials.

[0125] These figures visually reveal the spatial distribution differences of different demagnetization modes, verify the effectiveness and distinguishability of "magnetic density waveform area difference" as a fault feature, provide key data support for constructing a high-precision multi-class demagnetization fault classification model, and demonstrate the strong robustness and universality of this method in identifying multiple demagnetization types.

[0126] The above primarily describes the solutions of the embodiments of this application from the perspective of device implementation. It is understood that each device, for example, a multi-type demagnetization fault diagnosis device for a permanent magnet synchronous motor, includes at least one of the hardware structures and software modules corresponding to each function in order to achieve the above-mentioned functions. Those skilled in the art should readily recognize that, based on the units and model steps of the various examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is implemented in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0127] This application embodiment can divide a multi-type demagnetization fault diagnosis device for permanent magnet synchronous motors into functional units based on the above method example. For example, each function can be divided into its own functional unit, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or software. It should be noted that the unit division in this application embodiment is illustrative and only represents a logical functional division; other division methods may be used in actual implementation.

[0128] When using integrated units, Figure 15 A possible structural schematic diagram of a multi-type demagnetization fault diagnosis device for permanent magnet synchronous motors (referred to as demagnetization fault diagnosis device 50) involved in the above embodiments is shown. The demagnetization fault diagnosis device 50 includes a processing unit 501 and a communication unit 502, and may also include a storage unit 503. Figure 15 The schematic diagram shown can be used to illustrate the structure of a multi-type demagnetization fault diagnosis device for permanent magnet synchronous motors involved in the above embodiments.

[0129] when Figure 15 The schematic diagram shown illustrates the structure of a multi-type demagnetization fault diagnosis device for a permanent magnet synchronous motor involved in the above embodiments. The processing unit 501 is used to control and manage the operation of the multi-type demagnetization fault diagnosis device for a permanent magnet synchronous motor. The communication unit 502 is used for the multi-type demagnetization fault diagnosis device for a permanent magnet synchronous motor to communicate with other devices. The storage unit 503 is used to store the program code and data of the multi-type demagnetization fault diagnosis device for a permanent magnet synchronous motor.

[0130] The communication unit is used to acquire air gap magnetic flux density signals of the permanent magnet synchronous motor under different demagnetization faults; the demagnetization fault includes fault type and fault degree; the processing unit is used to acquire magnetic flux density difference signals based on the air gap magnetic flux density signals of the demagnetization fault and the reference air gap magnetic flux density signals; divide the waveform of the magnetic flux density difference signal into several waveform regions, calculate the area of ​​the waveform region corresponding to each magnetic pole, and obtain the air gap magnetic flux density waveform area difference; calculate the fault feature quantity of each magnetic pole based on the air gap magnetic flux density waveform area difference, and obtain a multi-dimensional fault feature vector; and construct a fault classification model based on the multi-dimensional feature vector and the corresponding fault type; and construct a demagnetization degree prediction model based on the air gap magnetic flux density signal and the corresponding fault degree.

[0131] For example, communication unit 502 is used to acquire air gap magnetic flux density signals of permanent magnet synchronous motor under different demagnetization faults; demagnetization faults include fault type and fault degree;

[0132] Processing unit 501 is used to obtain magnetic density difference signal based on air gap magnetic density signal of demagnetization fault and reference air gap magnetic density signal; divide the waveform of magnetic density difference signal into several waveform regions, calculate the area of ​​waveform region corresponding to each magnetic pole, and obtain the air gap magnetic density waveform area difference;

[0133] In one possible implementation, the processing unit 501 is further configured to calculate the fault characteristic quantity of each magnetic pole based on the area difference of the air gap magnetic flux density waveform to obtain a multidimensional fault feature vector; and to construct a fault classification model based on the multidimensional feature vector and the corresponding fault type; and to construct a demagnetization degree prediction model based on the air gap magnetic flux density signal and the corresponding fault degree.

[0134] The processing unit 501 can be a processor or a controller, and the communication unit 502 can be a communication interface, transceiver, transceiver circuit, transceiver device, etc. The term "communication interface" is a general term and may include one or more interfaces. The storage unit 503 can be a memory. When the demagnetization fault diagnosis device 50 is a chip, the processing unit 501 can be a processor or a controller, and the communication unit 502 can be an input interface and / or an output interface, pins, or circuits, etc. The storage unit 503 can be a storage unit within the chip (e.g., a register, cache, etc.) or a storage unit located outside the chip (e.g., read-only memory (ROM), random access memory (RAM, etc.).

[0135] The communication unit can also be called a transceiver unit. The antenna and control circuit with transceiver functions in the demagnetization fault diagnosis device 50 can be considered as the communication unit 502 of the demagnetization fault diagnosis device 50, and the processor with processing functions can be considered as the processing unit 501 of the demagnetization fault diagnosis device 50. Optionally, the device in the communication unit 502 that implements the receiving function can be considered as a communication unit, which is used to execute the receiving steps in the embodiments of this application. The communication unit can be a receiver, a receiver circuit, etc. The device in the communication unit 502 that implements the transmitting function can be considered as a transmitting unit, which is used to execute the transmitting steps in the embodiments of this application. The transmitting unit can be a transmitter, a transmitter, a transmitting circuit, etc.

[0136] Figure 15If the integrated units in the process are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. Storage media for storing computer software products include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0137] Figure 15 The units in the process can also be called modules; for example, a processing unit can be called a processing module.

[0138] This application embodiment also provides a hardware structure diagram of a multi-type demagnetization fault diagnosis device for permanent magnet synchronous motors (denoted as demagnetization fault diagnosis device 60), see [link to diagram]. Figure 16 The demagnetization fault diagnosis device 60 includes a processor 601, and optionally, a memory 602 connected to the processor 601.

[0139] In the first possible implementation, see Figure 16 The demagnetization fault diagnosis device 60 also includes a transceiver 603. The processor 601, memory 602, and transceiver 603 are connected via a bus. The transceiver 603 is used to communicate with other devices or communication networks. Optionally, the transceiver 603 may include a transmitter and a receiver. The device in the transceiver 603 that implements the receiving function can be considered as a receiver, which is used to perform the receiving steps in the embodiments of this application. The device in the transceiver 603 that implements the transmitting function can be considered as a transmitter, which is used to perform the transmitting steps in the embodiments of this application.

[0140] Based on the first possible implementation method Figure 16 The schematic diagram shown can be used to illustrate the structure of a multi-type demagnetization fault diagnosis device for permanent magnet synchronous motors involved in the above embodiments.

[0141] in, Figure 16 This can also be illustrated as a system chip in a multi-type demagnetization fault diagnosis device for permanent magnet synchronous motors. In this case, the actions performed by the aforementioned multi-type demagnetization fault diagnosis device for permanent magnet synchronous motors can be implemented by this system chip. The specific actions performed can be found above and will not be repeated here.

[0142] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

Claims

1. A method for diagnosing multiple types of demagnetization faults of a permanent magnet synchronous motor, characterized in that, The method comprises the following steps: Obtaining air gap magnetic flux signals of a permanent magnet synchronous motor under different demagnetization faults; The demagnetization fault includes a fault type and a fault degree; Based on the air gap magnetic flux signals of the demagnetization fault and the reference air gap magnetic flux signals, a magnetic flux difference signal is obtained; wherein the reference air gap magnetic flux signals are the air gap magnetic flux signals of the permanent magnet synchronous motor in a healthy state; The magnetic flux difference signal waveform is divided into a plurality of waveform regions, the area of each magnetic pole corresponding to the waveform region is calculated, and the air gap magnetic flux waveform area difference is obtained; Based on the air gap magnetic flux waveform area difference, the fault characteristic quantity of each magnetic pole is calculated, and a multi-dimensional fault feature vector is obtained; The failure characteristic quantity of each magnetic pole , comprising: ； wherein, is the effective length of the permanent magnet motor core, A is the area of the cross section of the permanent magnet core, i is the magnetic pole number, is the area difference of the air gap magnetic flux waveform under the i-th magnetic pole, t0 is the starting position of the time window, and t1 is the termination position of the time window. The area difference of the i-th magnetic pole lower air-gap magnetic flux density waveform comprising: ； Where j is the sampling point number, For the first Within the time window corresponding to the first magnetic pole The magnetic flux density difference at each sampling point For the first The angle interval corresponding to the nth sampling point, N is the nth sampling point. The total number of sampling points within the time window corresponding to each magnetic pole; Based on the multi-dimensional feature vector and the corresponding fault type, a fault classification model is constructed; wherein the fault classification model is constructed based on a PSO-SVM fault classification model; And based on the air gap magnetic flux signals and the corresponding fault degree, a demagnetization degree prediction model is constructed; wherein the demagnetization degree prediction model is constructed based on a CNN-Bilstm-Attention model.

2. The method according to claim 1, characterized in that, The method for obtaining air gap magnetic flux signals of a permanent magnet synchronous motor under different demagnetization faults comprises the following steps: Establishing a finite element model of the permanent magnet synchronous motor; According to the demagnetization principle of the permanent magnet, the coercive force of the permanent magnet in the finite element model of the permanent magnet synchronous motor is changed, the finite element model of different demagnetization faults is simulated, and the air gap magnetic flux signals of different demagnetization faults are extracted.

3. The method according to claim 1, characterized in that, The method for obtaining the magnetic flux difference signal comprises the following steps: Align the rotor position of the air gap magnetic flux signal to be diagnosed with the reference air gap magnetic flux signal, and subtract the amplitude of the corresponding rotor position to obtain the magnetic flux difference signal.

4. The method according to claim 1, characterized in that, The division method for dividing the magnetic flux difference signal waveform into a plurality of waveform regions comprises the following steps: A sliding window method is used to divide the magnetic flux difference signal in one mechanical cycle into a plurality of continuous and non-overlapping time windows, and a plurality of waveform regions are obtained; wherein the number of time windows is equal to the number of magnetic poles of the permanent magnet synchronous motor.

5. The method according to claim 1, characterized in that, The fault classification model is constructed based on a PSO-SVM fault classification model, which comprises the following steps: Extracting a plurality of multi-dimensional feature vectors and corresponding fault type labels under different fault states from historical data to construct a demagnetization fault sample library; The multi-dimensional feature vectors in the demagnetization sample library are used as the input data of the PSO-SVM fault classification model, and the fault type labels are used as the output data of the PSO-SVM fault classification model. The PSO-SVM fault classification model is trained through the input data and the output data, and a fault classification model is obtained.

6. The method according to claim 1, characterized in that, The CNN-BiLSTM-Attention model comprises a convolutional neural network layer, a bidirectional long short-term memory network layer, and an attention mechanism layer; The convolutional neural network layer is used to extract local features of the air gap magnetic flux signal; The bidirectional long short-term memory network layer is used to learn the time sequence dependence of the features processed by the convolutional neural network layer; The attention mechanism layer is used to assign different weights to the output sequence of the bidirectional long short-term memory network layer.

7. The method according to claim 6, characterized in that, The demagnetization degree prediction model is constructed based on the CNN-Bilstm-Attention model, which comprises the following steps: Air gap magnetic flux signal data of the permanent magnet synchronous motor under different demagnetization degrees is obtained from historical data, and the corresponding demagnetization degree true value is labeled; The air-gap magnetic flux density signal data is used as input data of the CNN-BiLSTM-Attention model, and the demagnetization degree true value is used as output data of the CNN-BiLSTM-Attention model, the CNN-BiLSTM-Attention model is trained through the input data and the output data, and a demagnetization degree prediction model is obtained.

8. A permanent magnet synchronous motor multi-type demagnetization fault diagnosis system, based on the permanent magnet synchronous motor multi-type demagnetization fault diagnosis method of any one of claims 1-7, characterized in that, The method comprises an acquisition module (100), a feature extraction module (200) and a fault model construction module (300); The acquisition module (100) is configured to acquire air-gap magnetic flux density signals of the permanent magnet synchronous motor under different demagnetization faults. The feature extraction module (200) is configured to acquire a magnetic flux difference signal based on the air-gap magnetic flux density signals and reference air-gap magnetic flux density signals of the demagnetization faults; divide the magnetic flux difference signal waveform into a plurality of waveform regions; calculate areas of the waveform regions corresponding to respective magnetic poles; and obtain air-gap magnetic flux waveform area differences. Based on the air-gap magnetic flux waveform area differences, fault feature quantities of the respective magnetic poles are calculated to obtain a multi-dimensional fault feature vector. The failure characteristic quantity of each magnetic pole , comprising: ； wherein, is the effective length of the permanent magnet motor core, A is the area of the cross section of the permanent magnet core, i is the magnetic pole number, is the area difference of the air gap magnetic flux waveform under the i-th magnetic pole, t0 is the starting position of the time window, and t1 is the termination position of the time window. The area difference of the i-th magnetic pole lower air-gap magnetic flux density waveform comprising: ； Where j is the sampling point number, For the first Within the time window corresponding to the first magnetic pole The magnetic flux density difference at each sampling point For the first The angle interval corresponding to the nth sampling point, N is the nth sampling point. The total number of sampling points within the time window corresponding to each magnetic pole; The fault model construction module (300) is configured to construct a fault classification model based on the multi-dimensional feature vector and a corresponding fault type; and construct a demagnetization degree prediction model based on the air-gap magnetic flux density signals and a corresponding demagnetization degree.

Images