Permanent magnet synchronous motor fault prediction method and early warning system

Abstract

The invention provides a permanent magnet synchronous motor fault prediction method and early warning system, and the method comprises the steps: collecting motor operation parameters through a sensor combination, and the motor operation parameters comprise an electric control bus voltage, an electric control bus current, a cell temperature, a motor rotation speed, and a motor torque; preprocessing motor operation parameters based on an MIV algorithm to screen key characteristic variables and construct a fault data set; constructing a BP neural network; optimizing a weight and a threshold value of the BP neural network based on an IPSO algorithm; performing data training on the BP neural network by using the optimized weight and threshold as initial values of the BP neural network to obtain a corresponding MIV-IPSO-BP neural network model; and inputting the real-time operation parameters to the MIV-IPSO-BP neural network model, and obtaining fault prediction data, so as to solve the technical problems that an existing method is very easy to cause misdiagnosis, and the signal acquisition process is very easy to be interfered by the external environment, and the fault prediction result is influenced.

Description

Technical Field

[0001] The present application relates to the technical field of motor fault prediction and identification methods, and specifically to a permanent magnet synchronous motor fault prediction method and early warning system. Background Art

[0002] Existing motor fault identification methods mostly extract features from vibration signals or current signals. However, there are many types of permanent magnet synchronous motor faults, including mechanical faults, electrical faults, electromagnetic faults, coupling faults, etc., and the causes of the faults are complex. Different fault types may lead to the same fault signal, and different fault signals may correspond to the same fault type.

[0003] In patent CN110969096A, the three-phase current and radial vibration data of the permanent magnet synchronous motor are first collected; then, the data is preprocessed, including time window division and normalization processing; then, model training is performed, the time domain and frequency domain characteristics of the processed data are calculated, and label matching is performed to train the neural network model; finally, real-time fault diagnosis is performed, the time domain and frequency domain characteristics of the processed data are calculated, the predicted data is formed, and the trained neural network model is input to obtain the electromechanical fault state of the motor.

[0004] In patent CN116953514A, the current data of the permanent magnet synchronous motor is converted into image data as the input of the deep neural network, and the deep convolutional network model is used to extract the PMSM fault features contained in the image. Finally, a classifier is used to identify and distinguish faults of different categories and degrees.

[0005] The current data-driven motor fault prediction and identification methods have high requirements on data quality. When dealing with large samples and high-dimensional data, they face challenges such as high computational complexity and difficulty in data processing and analysis. Therefore, the existing methods are prone to misdiagnosis. In addition, the process of collecting these signals is easily disturbed by the external environment, which affects the fault prediction results. Summary of the invention

[0006] The present application provides a permanent magnet synchronous motor fault prediction method and early warning system to solve the technical problem that the existing method is prone to misdiagnosis and is easily disturbed by the external environment during the process of collecting these signals, thus affecting the fault prediction results.

[0007] The present application provides a permanent magnet synchronous motor fault prediction method, comprising the steps of: The sensor combination is used to collect the motor operating parameters, including the electric control bus voltage, the electric control bus current, the battery core temperature, the motor speed, and the motor torque; Preprocess the motor operating parameters based on the MIV algorithm to screen key characteristic variables and build a fault data set; Construct BP neural network; Optimize the weights and thresholds of BP neural network based on IPSO algorithm; The optimized weights and thresholds are used as the initial values ​​of the BP neural network, and data training is performed on it to obtain the corresponding MIV-IPSO-BP neural network model; Input real-time operating parameters into the MIV-IPSO-BP neural network model and obtain fault prediction data.

[0008] Optionally, the step of preprocessing the motor operating parameters based on the MIV algorithm to screen key characteristic variables and construct a fault data set specifically includes the following steps: Cleaning motor operating parameters; Identify and handle outliers; Based on the MIV algorithm, the key characteristic variables are screened from the motor operating parameters after outlier processing; The screened key feature variables are normalized, and the transformation formula for normalization is: in, Represents the normalized input value; Input sample data processed by the MIV algorithm; and They represent the minimum and maximum values ​​in the data processed by the MIV algorithm.

[0009] Optionally, in the step of screening key characteristic variables of the motor operating parameters after abnormal value processing based on the MIV algorithm, the following steps are specifically included: The motor operating parameters after outlier processing are input into the MIV algorithm; Use the neural network model to train the MIV algorithm; Each independent variable in the motor operating parameters after abnormal value processing is added or subtracted by a certain amplitude based on its original value to form two new samples P1 and P2; Send P1 and P2 into the neural network model respectively to obtain two new simulation results Q1 and Q2; Take the difference between Q1 and Q2 to obtain the impact change value; Divide the impact change value by the total number of samples of the motor operating parameters after outlier processing, and obtain the MIV of the independent variable; Sort the independent variables according to their absolute value of MIV, and remove the independent variables that have a smaller impact on the dependent variable.

[0010] Optionally, the step of constructing a BP neural network specifically includes: The weights and thresholds of the BP neural network are set as the population parameters of the IPSO algorithm; The prediction error of the BP neural network is used as the fitness value of the IPSO algorithm; Calculate individual optimal values, global optimal values, and average optimal values; Update the particle swarm position, particle swarm velocity and the optimal position of each particle; Calculate the adaptive variance; Determine whether the adaptive variance is less than the preset range. If so, the particle swarm performs mutation operation. If not, calculate the fitness value.

[0011] Optionally, in the step of optimizing the weights and thresholds of the BP neural network based on the IPSO algorithm, the following steps are specifically included: Logistic chaotic mapping is used to initialize the population. The mathematical formula of Logistic mapping is as follows: {X}_{t+1}={X}_{t}\lambda \left ( {1-{X}_{t}} \right ), \lambda \in \left [ {0, 4} \right ], {X}_{t}\in \left ( {0, 1} \right ) in, Representative The population ratio of the generation, Representative The population ratio of the generation, represents the growth rate parameter; The Levy flight mechanism is used to adjust the jumping direction and step length to quickly approach the optimal solution. The update formula is as follows: ; in, Representative The population ratio of the generation, Representative The population ratio of the generation, Represents the step size scaling factor, with a value of 0.01. represents the Lévy random path, Represents an important parameter for adjusting Levi's flight stability, with a value range of .

[0012] Optionally, in the step of using the Levy flight mechanism to adjust the jumping direction and step length to quickly approach the optimal solution, the mathematical model of the Levy random path is: ; in, represents the Lévy random path, To adjust the important parameters of Levi's flight stability, and represents a variable that follows a normal distribution, represents the mathematical expectation, represents the standard gamma function, and represent and The variance of .

[0013] Optionally, in the step of initializing the population using Logistic chaotic mapping, When 3.57< At 4, the whole system is in a chaotic state.

[0014] Optionally, in the step of using the optimized weights and thresholds as initial values ​​of the BP neural network, performing data training on it, and obtaining the corresponding MIV-IPSO-BP neural network model, before performing data training, the following steps are also included: Determine the topological structure of the BP neural network. The input fault characteristics and output prediction characteristics correspond to the number of input layer nodes and the number of output layer nodes respectively. The number of hidden layer neurons is determined using the empirical formula: in, Represents the number of hidden layer nodes; Represents the number of input layer nodes, represents the number of output layer nodes, Represents any integer between 1 and 10.

[0015] Optionally, after the step of inputting the real-time operating parameters into the MIV-IPSO-BP neural network model and obtaining the fault prediction data, the method further includes the following steps: Perform residual analysis on the data obtained from fault prediction, and set the maximum residual as the warning threshold; Extract some data from the fault data set and import them into the MIV-IPSO-BP model for prediction to obtain the residual; Determine whether the residual is greater than the warning threshold. If so, trigger the alarm system.

[0016] Correspondingly, the present application also provides a permanent magnet synchronous motor fault warning system, including a memory and a processor, the memory being used to store executable program code; the processor being connected to the memory, and running a computer program corresponding to the executable program code by reading the executable program code, so as to execute the steps in the permanent magnet synchronous motor fault prediction method as mentioned above.

[0017] The present application provides a permanent magnet synchronous motor fault prediction method and early warning system, which adopts a variety of sensors to comprehensively collect the multi-dimensional operating parameters of the motor, such as the speed, torque, temperature, voltage, current, etc., combines big data processing technology, and uses the MIV algorithm to screen out characteristic variables with high correlation with faults, thereby realizing efficient integration and cleaning of massive heterogeneous data, and improving the accuracy and reliability of the prediction model.

[0018] The IPSO algorithm is combined with the BP neural network to optimize the network training process, improve the generalization ability and prediction accuracy of the model, and the algorithm can automatically learn and identify the precursor characteristics of equipment failures, and can respond quickly even to nonlinear, time-varying and complex failure modes.

[0019] The MIV-IPSO-BP neural network model enhances the optimization capability and convergence speed of the particle swarm algorithm, making the training process of the fault prediction model more efficient. This means that accurate prediction results can be obtained more quickly, improving the efficiency of engineering applications.

[0020] The MIV-IPSO-BP neural network model improvement and data processing method are used to enhance the adaptability and generalization ability of the prediction model to fault data under different working conditions, which enables the prediction model to better cope with complex and changeable engineering environments and improves its practicality and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 It is a flow chart of the permanent magnet synchronous motor fault prediction method provided in this application. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application. In addition, it should be understood that the specific implementation methods described herein are only used to illustrate and explain the present application and are not used to limit the present application. In the present application, unless otherwise stated, the directional words used, such as "up", "down", "left", and "right", generally refer to the up, down, left, and right of the device in actual use or working state, specifically the drawing direction in the accompanying drawings.

[0024] The present application provides a permanent magnet synchronous motor fault prediction method and early warning system, which are described in detail below. It should be noted that the description order of the following embodiments does not limit the preferred order of the embodiments of the present application. In the following embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0025] See also Figure 1 The current data-driven motor fault prediction and identification methods have very high requirements on data quality, and face challenges such as high computational complexity, data processing and analysis difficulties when processing large samples and high-dimensional data. The present application provides a permanent magnet synchronous motor fault prediction method, which discloses a permanent magnet synchronous motor fault diagnosis method based on MIV-IPSO-BP neural network. The method can not only process high-dimensional motor fault data, but also improve the generalization ability of the model, adapt to the variable factors in the motor operation process in a complex working environment, and improve the accuracy of fault prediction and identification.

[0026] See also Figure 1 , the permanent magnet synchronous motor fault prediction method specifically includes the following steps: S100, using a sensor combination to collect motor operating parameters, the motor operating parameters including electric control bus voltage, electric control bus current, battery core temperature, motor speed, and motor torque; A variety of sensors are used to realize real-time monitoring and data collection of motor operating parameters. The main parameters collected include electric control bus voltage, electric control bus current, battery core temperature, voltage temperature, motor speed, motor torque, etc. The data are stored in the database.

[0027] The sensor combination includes a voltage sensor, a current sensor, a temperature sensor, a speed sensor and a torque sensor, among which the voltage sensor is installed on the motor control circuit to measure the voltage of the motor in real time; the current sensor is installed in the circuit at the input and output ends of the motor to measure the current of the motor in real time; the temperature sensor is installed in the motor heat sink and around the battery cell to measure the temperature in real time; the speed sensor is installed on the rotating parts of the motor to measure the motor speed in real time; the torque sensor is installed on the transmission system of the motor to measure the torque output of the motor in real time.

[0028] A variety of sensors are used to comprehensively collect the motor's multi-dimensional operating parameters such as speed, torque, temperature, etc., combined with big data processing technology, and the MIV algorithm is used to screen out characteristic variables with a high correlation with faults, achieving efficient integration and cleaning of massive heterogeneous data, and improving the accuracy and reliability of the prediction model.

[0029] S200, preprocessing the motor operating parameters based on the MIV algorithm to screen key characteristic variables and construct a fault data set; When collecting motor data, due to interference from the external environment, especially electromagnetic interference and vibration noise, it is easy to cause signal quality to deteriorate, affecting the accuracy and reliability of data collection. Therefore, it is necessary to preprocess the data. The specific steps are as follows: S210, cleaning motor operating parameters; Use data cleaning techniques, such as smoothing filtering or moving average method, to remove noise and sudden outliers in the data to ensure the quality of the collected data is reliable.

[0030] S220, identifying and processing outliers; Perform outlier detection and use statistical methods or rule-based methods to identify and process outliers to prevent them from affecting subsequent analysis. The above statistical methods can include mean and standard deviation method, box plot method, and sliding average method. The rule-based methods include fixed threshold method and sliding window method.

[0031] S230, screening key characteristic variables of the motor operating parameters after abnormal value processing based on the MIV algorithm; The mean influence value algorithm (MIV) is considered to be one of the best indicators for evaluating the influence of independent variables on dependent variables through neural networks. MIV can be used to screen out characteristic variables with a high correlation with motor faults.

[0032] The steps of screening key characteristic variables of motor operating parameters after abnormal value processing based on the MIV algorithm specifically include the following steps: S231, inputting the motor operating parameters after abnormal value processing into the MIV algorithm; S232, using a neural network model to train the MIV algorithm; S233, adding or subtracting a certain amplitude from the original value of each independent variable in the motor operating parameters after abnormal value processing to form two new samples P1 and P2; S234, respectively sending P1 and P2 into the neural network model to obtain two new simulation results Q1 and Q2; S235, taking the difference between Q1 and Q2 to obtain the impact change value; S236, dividing the impact change value by the total number of samples of the motor operating parameters after abnormal value processing, and obtaining the MIV of the independent variable; S237, sort the MIV of each independent variable according to its absolute value, and remove the independent variables that have a small impact on the dependent variable; The independent variables with less influence on the dependent variable are removed according to the absolute value of MIV to achieve the purpose of data dimensionality reduction.

[0033] S240, normalizing the selected key feature variables, wherein the transformation formula of the normalization processing is: in, Represents the normalized input value; Input sample data processed by the MIV algorithm; and They represent the minimum and maximum values ​​in the data processed by the MIV algorithm.

[0034] In order to eliminate the dimensional differences between parameters and ensure that the influence of each eigenvector on the model is relatively uniform, the data needs to be normalized before training the neural network, that is, the data needs to be scaled to a standard range, which helps to improve the performance of the model and the stability of training.

[0035] S300, construct BP neural network; The training process of BP neural network includes initializing network parameters, performing forward propagation to calculate output, and then calculating errors through loss function. In the back propagation stage, the errors are used to calculate the gradients of weights and biases of each layer, and these parameters are updated through optimization algorithm. This process is repeated until the loss function converges or the set number of training rounds is reached. Finally, the validation set is used to evaluate the model performance, and the parameters are adjusted to optimize the model as needed.

[0036] The steps of constructing the BP neural network specifically include: S310, setting the weight and threshold of the BP neural network as the population parameters of the IPSO algorithm; S320, using the prediction error of the BP neural network as the fitness value of the IPSO algorithm; S330, calculating the individual optimal value, the global optimal value and the average optimal value; S340, updating the particle group position, particle group speed and the optimal position of each particle; S350, calculating the adaptive variance; S360, determining whether the adaptive variance is less than a preset range, if so, the particle swarm performs a mutation operation, if not, calculating the fitness value.

[0037] S400, optimize the weights and thresholds of BP neural network based on IPSO algorithm; The mathematical theoretical basis of the PSO algorithm is: {V}_{i}=\left [ {{\nu}_{i,1},{\nu}_{i,2},...,{\nu}_{i,d}} \right ] and{X}_{i}=\left [{{x}_{i,1},{x}_{i,2},...,{x}_{i,d}} \right ] Respectively In the dimensional space The speed and position of each particle. In each iteration, the fitness function value of each particle is evaluated to determine The individual optimal position of each particle at the moment and the optimal position found by the entire group , each particle in the group updates its speed and position according to the two optimal positions, as shown in the following formula: {\nu}_{i,j}\left ( {t+1} \right )={\nu}_{i,j}\left ( {t} \right )+{c}_{1}{r}_{1}\left [ {{p}_{i,j}-{x}_{i,j}(t)} \right ]+{c}_{2}{r}_{2}\left [ {{p}_{g,j}-{x}_{i,j}(t)} \right ] Where: is the number of iterative evolutions; For particles Current position in time; is the velocity of the particle at generation t; is the individual position of each particle, , is the acceleration constant, here we assume , which are used to adjust the scale of each particle flying towards the individual and global optimal particle running directions respectively; , is a random number uniformly distributed between 0 and 1. In addition, in order to reasonably restrict the movement of particles, the particle speed range is set to \left [ {{\nu}_{max}, {\nu}_{min}} \right ] , the position interval is \left [ {{x}_{max}, {x}_{min}} \right ] .

[0038] The basic PSO is easy to use, but it is prone to local optimality and convergence accuracy problems. Therefore, multiple strategies are needed to improve it.

[0039] The traditional IPSO algorithm uses randomly generated data to initialize the population, which can easily lead to uneven population distribution and affect the convergence speed of the algorithm. The use of Logistic mapping can enhance the randomness and diversity of particles, improve the convergence speed of the algorithm, and make it easier to search for the global optimal solution.

[0040] When the IPSO algorithm optimizes the BP neural network, the improved particle swarm algorithm (IPSO) is used to optimize the weights and thresholds of the BP neural network, solving the problem that the standard BP network is prone to fall into local optimality and difficulty in adjusting parameters.

[0041] The above-mentioned steps of optimizing the weights and thresholds of the BP neural network based on the IPSO algorithm specifically include the following steps: Logistic chaotic mapping is used to initialize the population. The mathematical formula of Logistic mapping is as follows: {X}_{t+1}={X}_{t}\lambda \left ( {1-{X}_{t}} \right ), \lambda \in \left [ {0, 4} \right ], {X}_{t}\in \left ( {0, 1} \right ) in, Representative The population ratio of the generation, Representative The population ratio of the generation, represents the growth rate parameter; When 3.57< At 4, the whole system is in a chaotic state.

[0042] The Levy flight mechanism is used to adjust the jumping direction and step length to quickly approach the optimal solution. The update formula is as follows: ; in, Representative The population ratio of the generation, Representative The population ratio of the generation, Represents the step size scaling factor, with a value of 0.01. represents the Lévy random path, Represents an important parameter for adjusting Levi's flight stability, with a value range of .

[0043] The mathematical model of the above Levy random path is: ; in, represents the Lévy random path, To adjust the important parameters of Levi's flight stability, and represents a variable that follows a normal distribution, represents the mathematical expectation, represents the standard gamma function, and represent and The variance of .

[0044] By introducing the Levy flight mechanism and utilizing the uncertainty of the Levy flight direction and step length to perturb the population position, the diversity of the algorithm is improved, the probability of the algorithm falling into the local optimum is reduced, and thus the convergence speed is improved.

[0045] S500, using the optimized weights and thresholds as initial values ​​of the BP neural network, training it with data, and obtaining a corresponding MIV-IPSO-BP neural network model; The MIV-IPSO-BP neural network model combines the IPSO algorithm with the BP neural network, optimizes the network training process, and improves the generalization ability and prediction accuracy of the model. The algorithm can automatically learn and identify the precursor characteristics of equipment failures, and can respond quickly even to nonlinear, time-varying and complex failure modes.

[0046] Determine the topological structure of the BP neural network. The input fault characteristics and output prediction characteristics correspond to the number of input layer nodes and the number of output layer nodes respectively. The number of hidden layer neurons is determined using the empirical formula: in, Represents the number of hidden layer nodes; Represents the number of input layer nodes, represents the number of output layer nodes, Represents any integer between 1 and 10.

[0047] After setting other parameters of the BP neural network and related parameters of the IPSO algorithm, the MIV-IPSO-BP neural network model is used to predict faults of synchronous motors. The generalization ability and prediction accuracy of the proposed model are judged by observing the degree of fit between the predicted value and the true value.

[0048] S600, input real-time operating parameters into the MIV-IPSO-BP neural network model, and obtain fault prediction data; S700, performing residual analysis on the data obtained from the fault prediction, and setting the obtained maximum residual as the warning threshold; S800, extract part of the data from the fault data set and import it into the MIV-IPSO-BP model for prediction to obtain the residual; S900: Determine whether the residual is greater than a warning threshold, and if so, trigger an alarm system.

[0049] The fault warning system uses residual analysis to determine the fault warning threshold and monitors the motor temperature in real time. Once signs of potential fault are detected, the warning mechanism is immediately triggered.

[0050] Compared with the prior art, this embodiment has the following advantages: Higher prediction accuracy: The patent uses the IPSO algorithm to optimize the BP neural network and the MIV algorithm for feature selection, thereby providing a more accurate prediction of synchronous motor faults. Compared with traditional methods, this technology can predict faults more accurately.

[0051] Faster convergence speed: The improved algorithm enhances the optimization ability and convergence speed of the particle swarm algorithm, making the training process of the fault prediction model more efficient. This means that accurate prediction results can be obtained faster, improving the efficiency of engineering applications.

[0052] Higher generalization ability: The patented algorithm improvements and data processing methods enhance the adaptability and generalization ability of the prediction model to fault data under different working conditions. This enables the prediction model to better cope with complex and changing engineering environments, improving practicality and reliability.

[0053] Solve the limitations of traditional methods: Standard BP neural networks are prone to local optimality and difficult to determine parameters. By optimizing the BP neural network through IPSO, the optimal weights and thresholds are found to solve these limitations.

[0054] The present application also provides a permanent magnet synchronous motor fault warning system, including a memory and a processor, the memory being used to store executable program code; the processor being connected to the memory, and running a computer program corresponding to the executable program code by reading the executable program code, so as to execute the steps in the aforementioned permanent magnet synchronous motor fault prediction method.

[0055] The above is a detailed introduction to the permanent magnet synchronous motor fault prediction method and early warning system provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for predicting permanent magnet synchronous motor faults, characterized in that: Includes steps: The sensor combination is used to collect the motor operating parameters, including the electric control bus voltage, the electric control bus current, the battery core temperature, the motor speed, and the motor torque; Preprocess the motor operating parameters based on the MIV algorithm to screen key characteristic variables and build a fault data set; Construct BP neural network; Optimize the weights and thresholds of BP neural network based on IPSO algorithm; The optimized weights and thresholds are used as the initial values ​​of the BP neural network, and data training is performed on it to obtain the corresponding MIV-IPSO-BP neural network model; Input real-time operating parameters into the MIV-IPSO-BP neural network model and obtain fault prediction data. Wherein, in the step of optimizing the weights and thresholds of the BP neural network based on the IPSO algorithm, the following steps are specifically included: Logistic chaotic mapping is used to initialize the population. The mathematical formula of Logistic mapping is as follows: {X}_{t+1}={X}_{t}\lambda \left ( {1-{X}_{t}} \right ),\lambda \in \left [ {0,4} \right ],{X}_{t}\in \left ( {0,1} \right ) in, Representative The population ratio of the generation, Representative The population ratio of the generation, represents the growth rate parameter; The Levy flight mechanism is used to adjust the jumping direction and step length to quickly approach the optimal solution. The update formula is as follows: ； in, Representative The population ratio of the generation, Representative The population ratio of the generation, Represents the step size scaling factor, with a value of 0.

01. represents the Lévy random path, Represents an important parameter for adjusting Levi's flight stability, with a value range of .

2. The permanent magnet synchronous motor fault prediction method according to claim 1, characterized in that: The step of preprocessing the motor operating parameters based on the MIV algorithm to screen key characteristic variables and construct a fault data set specifically includes the following steps: Cleaning motor operating parameters; Identify and handle outliers; Based on the MIV algorithm, the key characteristic variables are screened from the motor operating parameters after outlier processing; The screened key feature variables are normalized, and the transformation formula for normalization is: in, Represents the normalized input value; Input sample data processed by the MIV algorithm; and They represent the minimum and maximum values ​​in the data processed by the MIV algorithm.

3. The permanent magnet synchronous motor fault prediction method according to claim 2, characterized in that: The step of screening key characteristic variables of the motor operating parameters after abnormal value processing based on the MIV algorithm specifically includes the following steps: The motor operating parameters after outlier processing are input into the MIV algorithm; Use the neural network model to train the MIV algorithm; Each independent variable in the motor operating parameters after abnormal value processing is added or subtracted by a certain amplitude based on its original value to form two new samples P1 and P2; Send P1 and P2 into the neural network model respectively to obtain two new simulation results Q1 and Q2; Take the difference between Q1 and Q2 to obtain the impact change value; Divide the impact change value by the total number of samples of the motor operating parameters after outlier processing, and obtain the MIV of the independent variable; Sort the independent variables according to their absolute value of MIV, and remove the independent variables that have a smaller impact on the dependent variable.

4. The permanent magnet synchronous motor fault prediction method according to claim 1, characterized in that: The steps of constructing the BP neural network specifically include: The weights and thresholds of the BP neural network are set as the population parameters of the IPSO algorithm; The prediction error of the BP neural network is used as the fitness value of the IPSO algorithm; Calculate individual optimal values, global optimal values, and average optimal values; Update the particle swarm position, particle swarm velocity and the optimal position of each particle; Calculate the adaptive variance; Determine whether the adaptive variance is less than the preset range. If so, the particle swarm performs mutation operation. If not, calculate the fitness value.

5. The permanent magnet synchronous motor fault prediction method according to claim 1, characterized in that: In the step of using the Levy flight mechanism to adjust the jumping direction and step length to quickly approach the optimal solution, the mathematical model of the Levy random path is: ； in, represents the Lévy random path, To adjust the important parameters of Levi's flight stability, and represents a variable that follows a normal distribution, represents the mathematical expectation, represents the standard gamma function, and represent and The variance of .

6. The permanent magnet synchronous motor fault prediction method according to claim 1, characterized in that: In the step of initializing the population using Logistic chaotic mapping, When 3.57< At 4, the whole system is in a chaotic state.

7. The permanent magnet synchronous motor fault prediction method according to claim 1, characterized in that: In the step of using the optimized weights and thresholds as the initial values ​​of the BP neural network, performing data training on it, and obtaining the corresponding MIV-IPSO-BP neural network model, before performing data training, the following steps are also included: Determine the topological structure of the BP neural network. The input fault characteristics and output prediction characteristics correspond to the number of input layer nodes and the number of output layer nodes respectively. The number of hidden layer neurons is determined using the empirical formula: in, Represents the number of hidden layer nodes; Represents the number of input layer nodes, represents the number of output layer nodes, Represents any integer between 1 and 10.

8. The method for predicting permanent magnet synchronous motor faults according to claim 1, characterized in that: After the step of inputting the real-time operating parameters into the MIV-IPSO-BP neural network model and obtaining the fault prediction data, the following steps are also included: Perform residual analysis on the data obtained from fault prediction, and set the maximum residual as the warning threshold; Extract some data from the fault data set and import them into the MIV-IPSO-BP model for prediction to obtain the residual; Determine whether the residual is greater than the warning threshold. If so, trigger the alarm system.

9. A permanent magnet synchronous motor fault warning system, characterized in that: include: A memory for storing executable program codes; as well as A processor is connected to the memory, and runs a computer program corresponding to the executable program code by reading the executable program code to execute the permanent magnet synchronous motor fault prediction method as described in any one of claims 1-8.

Images