A method, system, device and medium for monitoring the operating state of a permanent magnet synchronous motor

Abstract

The embodiment of the application provides a kind of permanent magnet synchronous motor operating state monitoring method, the method comprises: obtaining the simulation and actual three-phase current data of permanent magnet synchronous motor under fault state, respectively to three-phase current data is carried out continuous wavelet transform, obtain simulation and actual three-phase current time-frequency diagram;Three-phase current time-frequency diagram is standardized and is handled and is averaged fusion, obtain simulation and actual comprehensive time-frequency diagram;Fractal information dimension algorithm is established, and fault diagnosis model is constructed based on the algorithm;Based on simulation comprehensive time-frequency diagram, fault diagnosis model is trained, and pre-training model is obtained;Based on actual comprehensive time-frequency diagram, pre-training model is fine-tuned training, and fine-tuned pre-training model is obtained;Fine-tuned pre-training model is based on actual comprehensive time-frequency diagram and outputs prediction result;According to result, the operating state of motor is identified, the complexity and diversity of comprehensive extraction fault data are realized, and the sensitivity of fault diagnosis model to fault feature is improved.

Description

Technical Field

[0001] This application relates to the field of motor control technology, and in particular to a method, system, device and medium for monitoring the operating status of a permanent magnet synchronous motor. Background Technology

[0002] Permanent magnet synchronous motors (PMSMs) occupy an important position in industrial automation, aerospace, and electric vehicles due to their high efficiency, precise control, and high power density. However, under long-term and complex operating environments, these motors are prone to electrical and mechanical faults, affecting performance stability and safety, and causing serious consequences. Therefore, developing a high-precision, safe, and reliable fault diagnosis system for PMSMs is crucial. This system must possess fault warning and early diagnosis capabilities to prevent motor damage and ensure operational safety.

[0003] With the development of artificial intelligence and the Industrial Internet of Things, motor fault diagnosis is generally achieved by collecting data during motor operation, such as current, voltage and vibration signals, and using advanced signal processing and neural network technology to realize real-time detection and prediction of motor faults.

[0004] However, in the existing technology, the data-driven permanent magnet synchronous motor fault diagnosis model requires a large amount of high-quality data for training to obtain good performance. However, due to the difficulty in collecting motor fault data and the lack of utilization of prior information, the data model ignores some important clues or patterns when extracting fault features, which causes the model to be unable to accurately extract features and identify faults when faced with new or complex faults. Summary of the Invention

[0005] The main objective of this application is to propose a method, system, device, and medium for monitoring the operating status of a permanent magnet synchronous motor. This aims to address the problem that existing data-driven fault diagnosis methods for permanent magnet synchronous motors ignore the influence of prior information on the model and cannot fully extract the complexity and diversity of fault data, thereby improving the sensitivity of the fault diagnosis model to fault characteristics.

[0006] To achieve the above objectives, a first aspect of this application proposes a method for monitoring the operating status of a permanent magnet synchronous motor, the method comprising:

[0007] Acquire three-phase current data of a permanent magnet synchronous motor under fault conditions, wherein the three-phase current data includes simulated three-phase current data and actual three-phase current data;

[0008] Continuous wavelet transform is performed on the simulated three-phase current data and the actual three-phase current data respectively to obtain the time-frequency diagram of the simulated three-phase current and the time-frequency diagram of the actual three-phase current.

[0009] The simulated three-phase current time-frequency diagram and the actual three-phase current time-frequency diagram are standardized and averaged and fused respectively to obtain the simulated comprehensive time-frequency diagram and the actual comprehensive time-frequency diagram.

[0010] A fractal information dimension algorithm is established, and a fault diagnosis model is constructed based on the fractal information dimension algorithm.

[0011] The fault diagnosis model is trained based on the simulated integrated time-frequency diagram to obtain a pre-trained model;

[0012] Based on the actual integrated time-frequency diagram, the pre-trained model is fine-tuned to obtain the fine-tuned pre-trained model.

[0013] The fine-tuned pre-trained model outputs prediction results based on the actual integrated time-frequency plot;

[0014] The operating status of the permanent magnet synchronous motor is identified based on the prediction results.

[0015] The method provided in the first aspect can solve the problem that existing data-driven fault diagnosis methods for permanent magnet synchronous motors ignore the influence of prior information on the model and cannot fully extract the complexity and diversity of fault data, thereby improving the sensitivity of the fault diagnosis model to fault characteristics.

[0016] In one possible implementation, acquiring the simulated three-phase current data includes:

[0017] Obtain the physical characteristic data of the permanent magnet synchronous motor, and construct a motor model based on the physical characteristic data;

[0018] The motor model is controlled to operate in a preset fault mode, and simulated three-phase current data of the motor model when operating in the preset fault mode is collected.

[0019] In one possible implementation, the continuous wavelet transform is achieved through the following formula:

[0020]

[0021] Where x(t) represents the current data, ψ(t) represents the wavelet function, and ψ * Let D represent the conjugate function of the wavelet function, E represent the scaling parameter, and W represent the translation parameter. x (D, E) represents the transformation results at scale D and position E;

[0022] The formula for the wavelet function is:

[0023]

[0024] Where ω0 represents the center frequency of the wavelet.

[0025] In one possible implementation, the standardization and averaging fusion of the simulated three-phase current time-frequency diagram and the actual three-phase current time-frequency diagram to obtain the simulated composite time-frequency diagram and the actual composite time-frequency diagram includes:

[0026] The simulated three-phase current time-frequency diagram and the actual three-phase current time-frequency diagram are standardized to obtain a standardized three-phase current time-frequency diagram. The standardization process is achieved through the following formula:

[0027]

[0028]

[0029]

[0030] Where D represents the scale parameter, E represents the translation parameter, and W' ia (D, E), W' ib (D, E) and W' ic (D, E) represent the normalized results of the time-frequency diagrams of phase a, phase b, and phase c currents at scale D and location E, respectively. ia (D, E), W ib (D, E) and W ic (D, E) represent the transformation results of phase a current data, phase b current data and phase c current data at scale D and position E, respectively;

[0031] The standardized three-phase current time-frequency diagrams are averaged and fused to obtain a comprehensive time-frequency diagram. The averaging and fusion are achieved using the following formula:

[0032]

[0033] Among them, W fu (D, E) represents the average fusion result of the standardized three-phase current time-frequency diagram at scale D and location E.

[0034] In one possible implementation, the fault diagnosis model includes: ResNet18 and three fully connected layers;

[0035] The fault diagnosis model constructed based on the fractal information dimension algorithm includes:

[0036] Construct a fractal information dimension embedding model based on the fractal information dimension algorithm;

[0037] The ResNet18 is selected as the backbone structure of the fault diagnosis model. The ResNet18 includes: an initial convolutional layer, an average pooling layer, and a convolutional network, wherein the convolutional network is composed of residual blocks.

[0038] The fractal information dimension embedding model is connected after the residual block, the output of the average pooling layer is converted into a vector, and the three fully connected layers are connected after the average pooling layer to obtain the fault diagnosis model.

[0039] The last fully connected layer of the three fully connected layers contains multiple neurons, each of which is used to represent the characteristics of the fault state or the normal state of the permanent magnet synchronous motor.

[0040] In one possible implementation, the construction of the fractal information dimension embedding model based on the fractal information dimension algorithm includes: fractal information dimension calculation, fractal information dimension expansion, and fractal information dimension fusion.

[0041] in,

[0042] The steps for calculating the fractal information dimension include: obtaining an original feature after convolution processing; calculating the probability dimension matrix of the original feature based on the fractal information dimension algorithm; and calculating the probability dimension matrix using the following formula:

[0043]

[0044] in, Let PD(·) represent the probability dimension matrix, and let PD(·) represent the fractal information dimension algorithm. This represents the original feature.

[0045] Based on the nearest neighbor interpolation algorithm, the fractal information dimension is obtained, which is achieved by the following formula:

[0046]

[0047] in, The fractal information dimension is represented by NN(·), and the nearest neighbor interpolation algorithm is represented by NN(·).

[0048] The step of expanding the fractal information dimension includes: performing an inner product between the fractal information dimension and the original feature to obtain the fractal information prior matrix. The inner product is achieved by the following formula:

[0049]

[0050] The mean of each channel is calculated using the following formula:

[0051]

[0052] Where, δ i express The mean of the i-th channel, x ijk express The mean matrix δ is obtained by calculating the mean value of each channel for the pixel values ​​in the j-th row and k-th column of the i-th channel.

[0053] The fractal information dimension fusion step includes: fusing the original features, the fractal information prior matrix, and the mean matrix to obtain the fractal information dimension embedding model. The fractal information dimension fusion is achieved through the following formula:

[0054]

[0055] in, λ represents the fractal information dimension embedding feature, and λ represents a hyperparameter used to adjust the ratio of the pixel value of each channel to the mean of the corresponding channel.

[0056] In one possible implementation, the step of pre-training the fault diagnosis model based on the simulated integrated time-frequency plot includes:

[0057] Set the number of training rounds, training batches, and pre-training learning rate;

[0058] The learning rate decay strategy is used to verify the first accuracy of the fault diagnosis model based on the simulated integrated time-frequency graph. Training is stopped when the first accuracy no longer improves, and the pre-trained model is obtained.

[0059] The steps for fine-tuning and pre-training the pre-trained model based on the actual integrated time-frequency graph include:

[0060] Freeze the neural network layer parameters in the pre-trained model, and fine-tune the three fully connected layers. The fine-tuning training steps include:

[0061] Set the number of fine-tuning training rounds and the fine-tuning training learning rate, wherein the fine-tuning training learning rate is less than the pre-training learning rate;

[0062] The learning rate decay strategy is used to verify the second accuracy of the fault diagnosis model based on the actual integrated time-frequency graph. Training is stopped when the second accuracy no longer improves, and the fine-tuned pre-trained model is obtained.

[0063] To achieve the above objectives, a second aspect of this application provides a fault diagnosis system for a permanent magnet synchronous motor, the system comprising:

[0064] Data acquisition module: used to acquire three-phase current data of permanent magnet synchronous motor under fault conditions, wherein the three-phase current data includes simulated three-phase current data and actual three-phase current data;

[0065] Continuous wavelet transform module: used to perform continuous wavelet transform on the simulated three-phase current data and the actual three-phase current data respectively to obtain the time-frequency diagram of the simulated three-phase current and the time-frequency diagram of the actual three-phase current;

[0066] Processing module: used to standardize and average the simulated three-phase current time-frequency diagram and the actual three-phase current time-frequency diagram respectively to obtain the simulated comprehensive time-frequency diagram and the actual comprehensive time-frequency diagram;

[0067] Algorithm building module: used to build fractal information dimension algorithms;

[0068] Model building module: used to build a fault diagnosis model based on the fractal information dimension algorithm;

[0069] Model pre-training module: used to pre-train the fault diagnosis model based on the simulation integrated time-frequency diagram to obtain a pre-trained model;

[0070] Model fine-tuning pre-training module: used to perform model fine-tuning pre-training on the pre-trained model based on the actual integrated time-frequency diagram, to obtain the fine-tuned pre-trained model;

[0071] Output module: used to output prediction results based on the actual integrated time-frequency diagram using the fine-tuned pre-trained model;

[0072] Diagnostic module: used to identify the operating status of the permanent magnet synchronous motor based on the prediction results.

[0073] Thirdly, an electronic device is provided, comprising a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the method for monitoring the operating status of a permanent magnet synchronous motor as described in any possible implementation of the first aspect.

[0074] Fourthly, a computer-readable storage medium is provided, the storage medium storing a computer program that, when executed by a processor, implements the method for monitoring the operating status of a permanent magnet synchronous motor as described in any possible implementation of the first aspect.

[0075] As can be seen from the technical solutions provided in one or more embodiments of this specification, the permanent magnet synchronous motor operation status monitoring method provided in this application acquires simulated three-phase current data and actual three-phase current data of the permanent magnet synchronous motor under fault conditions. Continuous wavelet transform is performed on the acquired three-phase current data to obtain simulated three-phase current time-frequency diagrams and actual three-phase current time-frequency diagrams. Then, the three-phase current time-frequency diagrams are standardized and averaged to obtain simulated comprehensive time-frequency diagrams and actual comprehensive time-frequency diagrams. A fractal information dimension algorithm is established, and a fault diagnosis model is constructed based on this algorithm. The fault diagnosis model is trained based on the simulated comprehensive time-frequency diagram to obtain a pre-trained model. The pre-trained model is then fine-tuned based on the actual comprehensive time-frequency diagram to obtain a fine-tuned pre-trained model. The fine-tuned pre-trained model outputs prediction results based on the actual comprehensive time-frequency diagram, and the operation status of the permanent magnet synchronous motor is identified based on the prediction results. The fault diagnosis model constructed based on the fractal information dimension algorithm can further capture subtle changes and complex structures in fault data. This solves the problem that existing data-driven fault diagnosis methods for permanent magnet synchronous motors ignore the influence of prior information on the model and cannot fully extract the complexity and diversity of fault data, thus improving the sensitivity of the fault diagnosis model to fault features. Attached Figure Description

[0076] To more clearly illustrate the technical solutions in one or more embodiments or prior art of this specification, the accompanying drawings used in the description of one or more embodiments or prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0077] Figure 1 This is a flowchart illustrating a method for monitoring the operating status of a permanent magnet synchronous motor, as provided in an embodiment of this application.

[0078] Figure 2 This is a schematic diagram of averaging and fusing the three-phase current time-frequency diagrams in a method for monitoring the operating status of a permanent magnet synchronous motor provided in this application embodiment.

[0079] Figure 3 This is a schematic diagram of the fractal information dimension algorithm in a method for monitoring the operating status of a permanent magnet synchronous motor provided in an embodiment of this application.

[0080] Figure 4 This is a schematic diagram illustrating the construction of a fractal information dimension embedding model based on a fractal information dimension algorithm in a method for monitoring the operating status of a permanent magnet synchronous motor provided in this application embodiment.

[0081] Figure 5This is a schematic diagram of the network structure of the fault diagnosis model in a method for monitoring the operating status of a permanent magnet synchronous motor provided in this application embodiment.

[0082] Figure 6 A structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0083] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described one or more embodiments are merely some embodiments of this specification, and not all embodiments. All other embodiments obtained by those skilled in the art based on one or more embodiments of this specification without creative effort should fall within the protection scope of this document.

[0084] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0085] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0086] Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other. The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0087] Figure 1 This is an optional flowchart of the method for monitoring the operating status of a permanent magnet synchronous motor provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S100 to S800.

[0088] Firstly, such as Figure 1 As shown, a method for monitoring the operating status of a permanent magnet synchronous motor is provided, the method comprising:

[0089] S100. Obtain the three-phase current data of the permanent magnet synchronous motor under fault conditions. The three-phase current data includes simulated three-phase current data and actual three-phase current data.

[0090] It should be noted that obtaining simulated three-phase current data of a permanent magnet synchronous motor under fault conditions includes, but is not limited to:

[0091] System Architecture: The system utilizes an NI (National Instruments) hardware platform, which includes NI CompactRIO, the NIPXI system, and the NI LabVIEW software environment. NI CompactRIO is used for acquiring and controlling real-time data; its high performance ensures the immediacy of data acquisition and the precise execution of control commands. The NIPXI system is used for processing high-performance data and simulation, supporting complex data analysis tasks and simulation environments with its superior computing power and flexibility. NI LabVIEW software serves as the system's development environment and user interface platform; its intuitive graphical programming and powerful integration capabilities simplify system design and debugging processes, improving simulation efficiency.

[0092] Building the simulation model framework: In NI LabVIEW Simulink, a high-fidelity motor model is built based on the motor's physical characteristics. This model includes the electromagnetic, mechanical, and control dynamics of the permanent magnet synchronous motor, accurately simulating its behavior under normal or fault conditions. Various fault modes are introduced into this model, such as:

[0093] Inter-turn short circuit: Inter-turn short circuit with a severity of 2%-20%, inter-turn short circuit with a severity of 20%-50%, inter-turn short circuit with a severity of 50%-80%; Phase-to-phase short circuit: AB phase short circuit, AC phase short circuit, BC phase short circuit, A phase-to-ground short circuit, B phase-to-ground short circuit, C phase-to-ground short circuit.

[0094] Configure the HIL (Hardware-in-the-Loop) hardware-in-the-loop simulation environment: Connect the NI CompactRIO module to the motor drive system and sensors to acquire real-time data on motor operation. The NIPXI system processes the signals collected from the NI CompactRIO module and performs real-time simulation of fault conditions. Configure the NI LabVIEW software to transmit and record the simulated three-phase current data of the motor under fault conditions in real time.

[0095] In the embodiments provided in this application, the HIL hardware-in-the-loop (HIL) simulation technology of NI devices is employed. This technology can accurately simulate and reproduce the operating states of motors under various fault conditions, which not only greatly enriches the diversity of data samples but also significantly increases the amount of usable data, providing solid support for the construction of the training set and enhancing the generalization ability of the machine learning model. This application does not limit the method for obtaining simulated three-phase current data of a permanent magnet synchronous motor under fault conditions.

[0096] It should also be noted that obtaining the actual three-phase current data of the permanent magnet synchronous motor under fault conditions includes, but is not limited to, using current sensors to collect the actual three-phase current data of the permanent magnet synchronous motor under fault conditions. This application does not limit the method for obtaining the actual three-phase current data of the permanent magnet synchronous motor under fault conditions.

[0097] S200. Perform continuous wavelet transform on the simulated three-phase current data and the actual three-phase current data respectively to obtain the time-frequency diagram of the simulated three-phase current and the time-frequency diagram of the actual three-phase current.

[0098] In this embodiment, continuous wavelet transform technology is introduced to convert simulated three-phase current data and actual three-phase current data into simulated three-phase current time-frequency diagrams and actual three-phase current time-frequency diagrams, which can effectively extract fault features while suppressing noise. This solves the problem that existing data-driven fault diagnosis methods for permanent magnet synchronous motors are easily affected by sensor accuracy and external environmental interference, resulting in significant noise in the collected data and affecting the accuracy of diagnosis.

[0099] S300: The simulated three-phase current time-frequency diagram and the actual three-phase current time-frequency diagram are standardized and averaged and fused to obtain the simulated comprehensive time-frequency diagram and the actual comprehensive time-frequency diagram.

[0100] S400. Establish a fractal information dimension algorithm and construct a fault diagnosis model based on the fractal information dimension algorithm.

[0101] For example, such as Figure 3 As shown, Figure 3 This diagram illustrates the fractal information dimension algorithm in a permanent magnet synchronous motor operation status monitoring method provided in this application embodiment. Typical fractal information dimension calculation methods are based on feature grayscale distribution and pixel block coverage methods, including but not limited to the following steps:

[0102] Feature partitioning: Three different scales are selected, denoted as ε1, ε2, and ε3 respectively. Let the feature be I with size N×N×M. At scale ε1, the feature is divided into a grid of size M / ε1×M / ε1, resulting in the feature... Among them, features At scales ε2 and ε3, features The grids are divided into two sizes, M / ε1ε2×M / ε1ε2 and M / ε1ε3×M / ε1ε3, respectively, and are represented by the following formulas:

[0103]

[0104]

[0105] in, These are represented as block representations at scales ε2 and ε3, respectively.

[0106] Calculation features Information dimension matrix: Let and The two blocks represent the features of the i-th layer, and are respectively denoted as and but The steps for calculating the information dimension matrix of a layer are as follows:

[0107] calculate The distribution function for each block at different scales is given by the following formula:

[0108]

[0109]

[0110] in, This represents the distribution function at scale ε². This represents the distribution function at scale ε3. express The grid in row j and column k.

[0111] Calculate based on the obtained distribution function The information entropy at scales ε2 and ε3 is specifically formulated as follows:

[0112]

[0113]

[0114] in, This represents the information entropy at scale ε². This represents the information entropy at scale ε3.

[0115] Based on the obtained information entropy, linear fitting is used to obtain... The information dimension is calculated using the following formula:

[0116]

[0117] in, express The information dimension.

[0118] Finally, calculate The information dimensions of all blocks in all feature layers are calculated and combined in their original order to obtain the fractal information dimension matrix ID. Simultaneously, to ensure that the fractal information dimension matrix conforms to the neural network parameter paradigm, a probabilistic mapping is performed on the fractal information dimension matrix. This probabilistic mapping is achieved through the following formula:

[0119]

[0120]

[0121] in, This represents the probability dimension matrix.

[0122] In this embodiment of the application, by establishing a fractal information dimension algorithm, the signal complexity and information content characteristics contained in the fault features can be quantified and transformed into a form that conforms to the neural network parameter paradigm, thereby enhancing the model's ability to extract prior information of fault features and improving the model's detection accuracy and generalization ability.

[0123] It should also be noted that when constructing a fault diagnosis model, using the fractal information dimension algorithm as a basis, the prior knowledge that characterizes the complexity and information content of fault characteristic signals is introduced into the neural network, which can enhance the model's ability to extract prior information about fault features.

[0124] S500: The fault diagnosis model is trained based on the simulation integrated time-frequency diagram to obtain a pre-trained model.

[0125] S600. Based on the actual integrated time-frequency diagram, the pre-trained model is fine-tuned to obtain the fine-tuned pre-trained model.

[0126] The S700 outputs prediction results based on the actual integrated time-frequency map by fine-tuning the pre-trained model.

[0127] S800 identifies the operating status of the permanent magnet synchronous motor based on the prediction results.

[0128] The method provided in the first aspect can solve the problem that existing data-driven fault diagnosis methods for permanent magnet synchronous motors ignore the influence of prior information on the model and cannot fully extract the complexity and diversity of fault data, thereby improving the sensitivity of the fault diagnosis model to fault characteristics.

[0129] In one possible approach, acquiring the simulated three-phase current data includes:

[0130] The physical characteristic data of the permanent magnet synchronous motor are obtained, and a motor model is constructed based on the physical characteristic data; the motor model is controlled to run in a preset fault mode, and the simulated three-phase current data of the motor model when running in the preset fault mode are collected.

[0131] For example, various fault modes can be introduced into the construction of the motor model, such as:

[0132] Inter-turn short circuits: inter-turn short circuits with a severity of 2%-20%, inter-turn short circuits with a severity of 20%-50%, and inter-turn short circuits with a severity of 50%-80%; Phase-to-phase short circuits: AB phase short circuit, AC phase short circuit, BC phase short circuit, A phase-to-ground short circuit, B phase-to-ground short circuit, and C phase-to-ground short circuit, for a total of 9 fault modes. The sampling frequency is set to 10kHz, and simulated three-phase current data is collected. This data is divided into training set and test set in a 7:3 ratio for training the fault diagnosis model.

[0133] In the embodiments provided in this application, the HIL hardware-in-the-loop (HIL) simulation technology of NI devices is employed. This technology can accurately simulate and reproduce the operating states of motors under various fault conditions, which not only greatly enriches the diversity of data samples but also significantly increases the amount of usable data, providing solid support for the construction of the training set and enhancing the generalization ability of the machine learning model. This application does not limit the method for obtaining simulated three-phase current data of a permanent magnet synchronous motor under fault conditions.

[0134] In one possible implementation, the continuous wavelet transform is achieved through the following formula:

[0135]

[0136] Where x(t) represents the current data, ψ(t) represents the wavelet function, and ψ * Let D represent the conjugate function of the wavelet function, E represent the scaling parameter, and W represent the translation parameter. x (D, E) represents the transformation result at scale D and position E.

[0137] The formula for the wavelet function is:

[0138]

[0139] Where ω0 represents the center frequency of the wavelet.

[0140] For example, the collected three-phase current data signals of the permanent magnet synchronous motor under fault conditions are respectively set as i a (t), i b (t) and i c (t), and then perform continuous wavelet transform to obtain the time-frequency diagrams of the currents in phases a, b, and c. The specific transform formulas are as follows:

[0141]

[0142]

[0143]

[0144] Where D represents the scale parameter, E represents the translation parameter, the scale parameter is set to 64, and the sampling frequency of the translation parameter and the motor are set to 10kHz. and These represent the transformation results of the current data of phase a, phase b, and phase c at scale D and location E, respectively.

[0145] This application incorporates continuous wavelet transform technology to convert simulated and actual three-phase current data into simulated and actual three-phase current time-frequency graphs, effectively extracting fault features while suppressing noise. This addresses the problem in existing data-driven permanent magnet synchronous motor fault diagnosis methods that are susceptible to interference from sensor accuracy and the external environment, resulting in significant noise in the collected data and affecting diagnostic accuracy.

[0146] like Figure 2 As shown, Figure 2 This diagram illustrates the averaging and fusion of three-phase current time-frequency diagrams in a method for monitoring the operating status of a permanent magnet synchronous motor provided in this application embodiment. In one possible implementation, the standardization and averaging fusion of the simulated three-phase current time-frequency diagram and the actual three-phase current time-frequency diagram to obtain a simulated composite time-frequency diagram and an actual composite time-frequency diagram includes:

[0147] The simulated three-phase current time-frequency diagram and the actual three-phase current time-frequency diagram are standardized to obtain a standardized three-phase current time-frequency diagram. The standardization process is achieved through the following formula:

[0148]

[0149]

[0150]

[0151] Where D represents the scale parameter, E represents the translation parameter, and W' ia (D, E), W' ib (D, E) and W' ic (D, E) represent the normalized results of the time-frequency diagrams of phase a, phase b, and phase c currents at scale D and location E, respectively. ia (D, E), W ib (D, E) and W ic (D, E) represent the transformation results of phase a current data, phase b current data, and phase c current data at scale D and position E, respectively.

[0152] The standardized three-phase current time-frequency diagrams are averaged and fused to obtain a comprehensive time-frequency diagram. The averaging and fusion are achieved using the following formula:

[0153]

[0154] Among them, W fu (D, E) represents the average fusion result of the standardized three-phase current time-frequency diagram at scale D and location E.

[0155] It should be noted that the integrated time-frequency diagram in the above embodiments takes into account the comprehensive information of the three-phase current, so it can fully reflect the operating status and potential fault characteristics of the motor, which helps to more accurately identify the type of motor fault.

[0156] like Figure 5 As shown, in one possible implementation, the fault diagnosis model includes ResNet18 and three fully connected layers.

[0157] The fault diagnosis model constructed based on the fractal information dimension algorithm includes:

[0158] A fractal information dimension embedding model is constructed based on the fractal information dimension algorithm; ResNet18 is selected as the backbone structure of the fault diagnosis model, wherein ResNet18 includes: an initial convolutional layer, an average pooling layer, and a convolutional network, wherein the convolutional network is composed of residual blocks; the fractal information dimension embedding model is connected after the residual blocks, the output of the average pooling layer is converted into a vector, and the three fully connected layers are connected after the average pooling layer to obtain the fault diagnosis model; wherein the last fully connected layer of the three fully connected layers contains multiple neurons, each neuron being used to represent the features of the fault state or the features of the normal state of the permanent magnet synchronous motor.

[0159] For example, the ResNet18 architecture consists of three main parts: the initial convolutional layer, the pooling layer, and the network core. The initial convolutional layer uses 7x7 kernels with a stride of 2, outputting 64 feature maps. This is followed by a 3x3 max-pooling layer with the same stride of 2 to further reduce the spatial size of the feature maps for initial feature extraction and dimensionality reduction. The network core contains convolutional networks, which in turn perform batch normalization and ReLU activation. Each residual block directly adds the input to the output of the convolutional layer via skip connections. This design allows the network to learn the residual between the input and output, helping to mitigate the vanishing gradient problem in deep networks. The number of kernels starts at 64 and doubles with each subsequent block. An additional convolutional layer precedes the residual block at the beginning of the network, followed by an average pooling layer. The four residual blocks of ResNet18 can be viewed as four convolutional networks with different structures, with fractal dimensionality embedding performed after each convolutional network. The number of channels in the four groups of residual blocks gradually increases: 64 channels in the first group, 128 channels in the second, 256 channels in the third, and 512 channels in the fourth. In the average pooling layer, ResNet18 compresses the spatial dimension of the feature map for each channel into a single value, transforming it into a fixed-length vector of 512. This vector is then connected to three fully connected layers to obtain the fault diagnosis model. These three fully connected layers contain 256, 128, and 10 neurons respectively, and are processed using the ReLU activation function. The number of nodes in the first two fully connected layers decreases sequentially to gradually refine the features and map the high-dimensional data. The last fully connected layer contains 10 neurons, each corresponding to a specific output category. These categories include nine different fault features and a neuron representing the normal operating state of the motor. Since the system health status of the motor needs to be monitored in real time, it is necessary to keep the model's complexity and parameters within a relatively small range; therefore, ResNet18 was chosen as the backbone structure of the fault diagnosis model.

[0160] In one possible implementation, such as Figure 4 As shown, Figure 4 This diagram illustrates the construction of a fractal information dimension embedding model based on a fractal information dimension algorithm in a method for monitoring the operating status of a permanent magnet synchronous motor provided in this application embodiment. The construction of the fractal information dimension embedding model based on the fractal information dimension algorithm includes: fractal information dimension calculation, fractal information dimension expansion, and fractal information dimension fusion.

[0161] The step of calculating the fractal information dimension includes: obtaining an original feature after convolution processing; calculating the probability dimension matrix of the original feature based on the fractal information dimension algorithm; and calculating the probability dimension matrix using the following formula:

[0162]

[0163] in, Let PD(·) represent the probability dimension matrix, and let PD(·) represent the fractal information dimension algorithm. This refers to the original feature.

[0164] It should be noted that the original feature

[0165] Based on the nearest neighbor interpolation algorithm, the fractal information dimension is obtained, which is achieved by the following formula:

[0166]

[0167] in, The fractal information dimension is represented by NN(·), and the nearest neighbor interpolation algorithm is represented by NN(·).

[0168] It should be noted that,

[0169] The step of expanding the fractal information dimension includes: performing an inner product between the fractal information dimension and the original feature to obtain the fractal information prior matrix. The inner product is achieved by the following formula:

[0170]

[0171] The mean of each channel is calculated using the following formula:

[0172]

[0173] Where, δ i express The mean of the i-th channel, x ijk express The mean matrix δ is obtained by calculating the mean value of each channel, based on the pixel values ​​in the j-th row and k-th column of the i-th channel.

[0174] The fractal information dimension fusion step includes: fusing the original features, the fractal information prior matrix, and the mean matrix to obtain the fractal information dimension embedding model. The fractal information dimension fusion is achieved through the following formula:

[0175]

[0176] in, λ represents the fractal information dimension embedding feature, and λ represents a hyperparameter used to adjust the ratio of the pixel value of each channel to the mean of the corresponding channel.

[0177] It should be noted that by establishing a fractal information dimension algorithm, the signal complexity and information content characteristics contained in fault features can be quantified and transformed into a form that conforms to the neural network parameter paradigm. This enhances the model's ability to extract prior information about fault features, thereby improving the model's detection accuracy and generalization ability. Based on the fractal information dimension algorithm, a fractal information dimension embedding model can be constructed to build a fault diagnosis model. This allows prior knowledge representing the signal complexity and information content of fault characteristics to be introduced into the neural network, thereby enhancing the model's ability to extract prior information about fault features.

[0178] In one possible implementation, the steps of pre-training the fault diagnosis model based on the simulated integrated time-frequency graph include: setting a rated number of training rounds, training batches, and a pre-training learning rate; using a learning rate decay strategy to verify the first accuracy of the fault diagnosis model based on the simulated integrated time-frequency graph; stopping training when the first accuracy no longer improves, thereby obtaining the pre-trained model.

[0179] For example, when the fault diagnosis model is pre-trained based on the simulated integrated time-frequency graph, the training epochs are set to 500, the batch size is set to 128, the pre-training learning rate is set to 0.00001, and a learning rate decay strategy is used. Every time the training dataset passes through the neural network 100 times, that is, after every 100 epochs, the learning rate is decayed back to the original 0.9 to avoid oscillations in the later stages of training. Then, the first accuracy of the model based on the simulated integrated time-frequency graph is verified. Training is stopped when the first accuracy no longer improves.

[0180] The steps of fine-tuning the pre-trained model based on the actual integrated time-frequency graph include: freezing the neural network layer parameters in the pre-trained model, fine-tuning the three fully connected layers, and the fine-tuning training steps include: setting the fine-tuning training rounds and the fine-tuning training learning rate, wherein the fine-tuning training learning rate is less than the pre-training learning rate; using a learning rate decay strategy to verify the second accuracy of the fault diagnosis model based on the actual integrated time-frequency graph, and stopping training when the second accuracy no longer improves, thereby obtaining the fine-tuned pre-trained model.

[0181] For example, after the pre-trained model is trained, to improve its adaptability to real-world application scenarios and ensure a high fault identification rate for real-world faults, the pre-trained model is fine-tuned based on the actual integrated time-frequency diagram. During this fine-tuning pre-training, the neural network parameters of the fractal information dimension network are frozen, and fine-tuning is performed only on the three fully connected layers to avoid overfitting. For example, if the fine-tuning training epochs are set to 20, and the fine-tuning learning rate is set to 10% of the pre-training learning rate, a learning rate decay strategy is used to obtain the fine-tuned pre-trained model, ensuring that the model can stably adapt to the actual three-phase current time-frequency diagram.

[0182] To achieve the above objectives, a second aspect of this application provides a fault diagnosis system for a permanent magnet synchronous motor, the system comprising:

[0183] Data acquisition module: used to acquire three-phase current data of permanent magnet synchronous motor under fault conditions, wherein the three-phase current data includes simulated three-phase current data and actual three-phase current data.

[0184] It should be noted that the data acquisition module employs HIL (Hardware-In-the-Loop) semi-physical simulation technology. This technology can accurately simulate and reproduce the operating states of motors under various fault conditions, greatly enriching the diversity of data samples and significantly increasing the amount of usable data. This provides solid support for the construction of the training set and enhances the generalization ability of the machine learning model. This application does not limit the method for acquiring simulated three-phase current data of a permanent magnet synchronous motor under fault conditions.

[0185] Continuous wavelet transform module: used to perform continuous wavelet transform on the simulated three-phase current data and the actual three-phase current data respectively to obtain the simulated three-phase current time-frequency diagram and the actual three-phase current time-frequency diagram.

[0186] In this embodiment, continuous wavelet transform technology is introduced into the continuous wavelet transform module to convert simulated three-phase current data and actual three-phase current data into simulated three-phase current time-frequency diagrams and actual three-phase current time-frequency diagrams, which can effectively extract fault features while suppressing noise. This solves the problem in existing data-driven permanent magnet synchronous motor fault diagnosis methods that are easily affected by sensor accuracy and external environmental interference, resulting in significant noise in the collected data and affecting the diagnostic accuracy.

[0187] Processing module: used to standardize and average the simulated three-phase current time-frequency diagram and the actual three-phase current time-frequency diagram respectively, to obtain the simulated comprehensive time-frequency diagram and the actual comprehensive time-frequency diagram.

[0188] Algorithm creation module: Used to create fractal information dimension algorithms.

[0189] It should be noted that the algorithm building module can quantify the signal complexity and information content characteristics contained in fault features by establishing a fractal information dimension algorithm, and transform them into a form that conforms to the neural network parameter paradigm, thereby enhancing the model's ability to extract prior information of fault features and improving the model's detection accuracy and generalization ability.

[0190] Model building module: used to build a fault diagnosis model based on the fractal information dimension algorithm.

[0191] Among them, the model building module constructs a fault diagnosis model based on the fractal information dimension algorithm, which can introduce prior knowledge that characterizes the complexity and information content of fault characteristic signals into the neural network, thereby enhancing the model's ability to extract prior information on fault features.

[0192] Model pre-training module: used to pre-train the fault diagnosis model based on the simulation integrated time-frequency diagram to obtain a pre-trained model.

[0193] Model fine-tuning pre-training module: used to perform model fine-tuning pre-training on the pre-trained model based on the actual integrated time-frequency diagram, to obtain the fine-tuned pre-trained model.

[0194] Output module: Used to output prediction results based on the actual integrated time-frequency diagram using the fine-tuned pre-trained model.

[0195] Diagnostic module: used to identify the operating status of the permanent magnet synchronous motor based on the prediction results.

[0196] The system provided in the second aspect can solve the problem that existing data-driven fault diagnosis methods for permanent magnet synchronous motors ignore the influence of prior information on the model and cannot fully extract the complexity and diversity of fault data, thereby improving the sensitivity of the fault diagnosis model to fault characteristics.

[0197] This application also provides an electronic device, such as... Figure 6 As shown, the electronic device 1400 includes:

[0198] One or more processors 1410;

[0199] The memory 1420 stores one or more programs, which, when executed by one or more processors 1410, enable the one or more processors 1410 to implement the permanent magnet synchronous motor operation status monitoring method provided in any embodiment of this application.

[0200] Memory 1420, as a non-transitory network system, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory 1420 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 1420 may optionally include remotely located memories 1420 relative to processor 1410, which can be connected to processor 1410 via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0201] The memory 1420 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1420 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1420 and is called and executed by the processor 1410.

[0202] The processor 1410 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0203] In some embodiments, the electronic device further includes:

[0204] Input / output interfaces are used to implement information input and output;

[0205] The communication interface is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0206] The bus transmits information between various components of the device (e.g., processor 1410, memory 1420, input / output interface, and communication interface);

[0207] The processor 1410, memory 1420, input / output interface, and communication interface can communicate with each other within the device via a bus.

[0208] An embodiment of this application also provides a computer-readable storage medium storing computer-executable instructions for executing the operation status monitoring method for a permanent magnet synchronous motor provided in any embodiment of this application.

[0209] An embodiment of this application also provides a computer program product, including a computer program or computer instructions, which are stored in a computer-readable storage medium. A processor of a computer device reads the computer program or computer instructions from the computer-readable storage medium and executes the computer program or computer instructions, causing the computer device to perform the method for monitoring the operating status of a permanent magnet synchronous motor provided in any embodiment of this application.

[0210] The system architecture and application scenarios described in the embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. Those skilled in the art will know that as the system architecture evolves and new application scenarios emerge, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0211] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0212] It will be understood by those skilled in the art that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0213] The above description, with reference to the accompanying drawings, illustrates some embodiments of this application, but does not limit the scope of the invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and spirit of this invention should be considered within the scope of this application.

[0214] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0215] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0216] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0217] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for monitoring the operating status of a permanent magnet synchronous motor, characterized in that, The method includes: Acquire three-phase current data of a permanent magnet synchronous motor under fault conditions, wherein the three-phase current data includes simulated three-phase current data and actual three-phase current data; Continuous wavelet transform is performed on the simulated three-phase current data and the actual three-phase current data respectively to obtain the time-frequency diagram of the simulated three-phase current and the time-frequency diagram of the actual three-phase current. The simulated three-phase current time-frequency diagram and the actual three-phase current time-frequency diagram are standardized and averaged and fused respectively to obtain the simulated comprehensive time-frequency diagram and the actual comprehensive time-frequency diagram. A fractal information dimension algorithm is established, and a fault diagnosis model is constructed based on the fractal information dimension algorithm. The fault diagnosis model is trained based on the simulated integrated time-frequency diagram to obtain a pre-trained model; Based on the actual integrated time-frequency diagram, the pre-trained model is fine-tuned to obtain the fine-tuned pre-trained model. The fine-tuned pre-trained model outputs prediction results based on the actual integrated time-frequency plot; The operating status of the permanent magnet synchronous motor is identified based on the prediction results.

2. The method according to claim 1, characterized in that, The acquisition of the simulated three-phase current data includes: Obtain the physical characteristic data of the permanent magnet synchronous motor, and construct a motor model based on the physical characteristic data; The motor model is controlled to operate in a preset fault mode, and simulated three-phase current data of the motor model when operating in the preset fault mode is collected.

3. The method according to claim 1, characterized in that, The continuous wavelet transform is achieved through the following formula: ， in, Represents current data. Describing wavelet functions, This represents the conjugate function of the wavelet function. Indicates the scale parameter. Indicates the translation parameter. Indicated in scale and location The transformation result on; The formula for the wavelet function is: ， in, This represents the center frequency of the wavelet.

4. The method according to claim 1, characterized in that, The standardization and averaging fusion of the simulated three-phase current time-frequency diagram and the actual three-phase current time-frequency diagram to obtain the simulated composite time-frequency diagram and the actual composite time-frequency diagram includes: The simulated three-phase current time-frequency diagram and the actual three-phase current time-frequency diagram are standardized to obtain a standardized three-phase current time-frequency diagram. The standardization process is achieved through the following formula: ， ， ， in, Indicates the scale parameter. Indicates the translation parameter. , and The time-frequency diagrams of phase a current, phase b current, and phase c current are respectively represented at different scales. and location The standardized processing results on the above , and These represent the phase a current data, phase b current data, and phase c current data at different scales. and location The transformation result on; The standardized three-phase current time-frequency diagrams are averaged and fused to obtain a comprehensive time-frequency diagram. The averaging and fusion are achieved using the following formula: ， in, This indicates that the standardized three-phase current time-frequency diagram is on a scale of and location The average fusion result.

5. The method according to claim 1, characterized in that, The fault diagnosis model includes: ResNet18 and three fully connected layers; The fault diagnosis model constructed based on the fractal information dimension algorithm includes: Construct a fractal information dimension embedding model based on the fractal information dimension algorithm; The ResNet18 is selected as the backbone structure of the fault diagnosis model. The ResNet18 includes: an initial convolutional layer, an average pooling layer, and a convolutional network, wherein the convolutional network is composed of residual blocks. The fractal information dimension embedding model is connected after the residual block, the output of the average pooling layer is converted into a vector, and the three fully connected layers are connected after the average pooling layer to obtain the fault diagnosis model. The last fully connected layer of the three fully connected layers contains multiple neurons, each of which is used to represent the characteristics of the fault state or the normal state of the permanent magnet synchronous motor.

6. The method according to claim 5, characterized in that, The construction of the fractal information dimension embedding model based on the fractal information dimension algorithm includes: fractal information dimension calculation, fractal information dimension expansion, and fractal information dimension fusion. in, The steps for calculating the fractal information dimension include: obtaining an original feature after convolution processing; calculating the probability dimension matrix of the original feature based on the fractal information dimension algorithm; and calculating the probability dimension matrix using the following formula: ， in, Denotes the probability dimension matrix. This represents the fractal information dimension algorithm. This represents the original feature. Based on the nearest neighbor interpolation algorithm, the fractal information dimension is obtained, which is achieved by the following formula: ， in, This represents the dimension of the fractal information. This refers to the nearest neighbor interpolation algorithm; The step of expanding the fractal information dimension includes: performing an inner product between the fractal information dimension and the original feature to obtain the fractal information prior matrix. The inner product is achieved by the following formula: ; The mean of each channel is calculated using the following formula: ， in, express The mean of the i-th channel, express The mean matrix is ​​obtained by calculating the mean value of the pixel values ​​in the j-th row and k-th column of the i-th channel. ; The fractal information dimension fusion step includes: fusing the original features, the fractal information prior matrix, and the mean matrix to obtain the fractal information dimension embedding model. The fractal information dimension fusion is achieved through the following formula: ， in, The embedding feature represents the dimension of fractal information. This represents a hyperparameter used to adjust the ratio of the pixel value of each channel to the mean value of that channel.

7. The method according to claim 5, characterized in that, The steps for pre-training the fault diagnosis model based on the simulated integrated time-frequency graph include: Set the number of training rounds, training batches, and pre-training learning rate; The learning rate decay strategy is used to verify the first accuracy of the fault diagnosis model based on the simulated integrated time-frequency graph. Training is stopped when the first accuracy no longer improves, and the pre-trained model is obtained. The steps for fine-tuning and pre-training the pre-trained model based on the actual integrated time-frequency graph include: Freeze the neural network layer parameters in the pre-trained model, and fine-tune the three fully connected layers. The fine-tuning training steps include: Set the number of fine-tuning training rounds and the fine-tuning training learning rate, wherein the fine-tuning training learning rate is less than the pre-training learning rate; The learning rate decay strategy is used to verify the second accuracy of the fault diagnosis model based on the actual integrated time-frequency graph. Training is stopped when the second accuracy no longer improves, and the fine-tuned pre-trained model is obtained.

8. A fault diagnosis system for a permanent magnet synchronous motor, characterized in that, The system includes: Data acquisition module: used to acquire three-phase current data of permanent magnet synchronous motor under fault conditions, wherein the three-phase current data includes simulated three-phase current data and actual three-phase current data; Continuous wavelet transform module: used to perform continuous wavelet transform on the simulated three-phase current data and the actual three-phase current data respectively to obtain the time-frequency diagram of the simulated three-phase current and the time-frequency diagram of the actual three-phase current; Processing module: used to standardize and average the simulated three-phase current time-frequency diagram and the actual three-phase current time-frequency diagram respectively to obtain the simulated comprehensive time-frequency diagram and the actual comprehensive time-frequency diagram; Algorithm building module: used to build fractal information dimension algorithms; Model building module: used to build a fault diagnosis model based on the fractal information dimension algorithm; Model pre-training module: used to pre-train the fault diagnosis model based on the simulation integrated time-frequency diagram to obtain a pre-trained model; Model fine-tuning pre-training module: used to perform model fine-tuning pre-training on the pre-trained model based on the actual integrated time-frequency diagram, to obtain the fine-tuned pre-trained model; Output module: used to output prediction results based on the actual integrated time-frequency diagram using the fine-tuned pre-trained model; Diagnostic module: used to identify the operating status of the permanent magnet synchronous motor based on the prediction results.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the method for monitoring the operating status of the permanent magnet synchronous motor according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for monitoring the operating status of the permanent magnet synchronous motor as described in any one of claims 1 to 7.

Images