Permanent magnet synchronous motor demagnetization fault positioning method based on branch current

Abstract

The invention discloses a demagnetization fault positioning method for a permanent magnet synchronous motor based on branch current, which comprises the following steps of: firstly, establishing a finite element model for motor simulation, obtaining a contrast data set of three-phase current and branch current for training from the finite element model, and then constructing a generative adversarial network for constructing a mapping relationship between the three-phase current and the branch current; the permanent magnet fault diagnosis method based on the branch current in combination with the intelligent algorithm is simple in application and high in accuracy, so that the fault positioning diagnosis of the permanent magnet synchronous motor does not need to install additional sensors or parts, the motor is prevented from being influenced by the additional sensors or parts during normal operation, and the reliability of the motor is improved. The method has the advantages of intelligence, adaptability, generalization, no influence on normal operation of the motor and the like.

Description

Technical Field

[0001] This invention belongs to the field of motor fault diagnosis technology, specifically relating to a method for locating demagnetization faults in permanent magnet synchronous motors based on branch current. Background Technology

[0002] PMSMs (Permanent Magnet Synchronous Motors) are widely used in sustainable energy fields such as wind power generation, new energy vehicles, and rail transportation due to their advantages of simple excitation, compact structure, and high power density. However, various faults inevitably occur during the manufacturing, assembly, and operation of these motors, such as rotor eccentricity, permanent magnet demagnetization, circuit faults, and bearing failures, which will seriously compromise the reliability and safety of motor operation. Permanent magnet demagnetization refers to the phenomenon of partial or uniform demagnetization due to the combined effects of the armature. Irreversible demagnetization faults are common in permanent magnet motors, and if not identified and corrected in time, they can cause serious equipment damage and property loss. Therefore, identifying and locating demagnetization faults in PMSMs is of great significance. Early detection and location of demagnetization faults can improve motor reliability and maintainability, extend motor lifespan, and reduce unexpected downtime and related losses. More importantly, understanding the type of motor fault and the location and number of faulty magnetic poles can help maintenance personnel perform accurate and timely maintenance. Therefore, research on PMSM demagnetization fault detection and fault location is of great importance.

[0003] Demagnetization fault refers to the partial or complete loss of the ability of permanent magnets in a motor rotor to generate magnetic flux or magnetic field. The main factor causing irreversible demagnetization of the magnets is the continuous operation of the motor under high temperature and overload conditions. When the motor is running, the permanent magnets in the rotor induce an electromotive force (EMF) in the stator windings. Under normal conditions, the induced EMF generated in each stator slot changes periodically. When a partial demagnetization fault occurs, the induced EMF generated in a single slot becomes irregular. When a demagnetized permanent magnet acts on a specific slot, the induced EMF decreases, while the induced EMF generated by other undemagnetized permanent magnets acting on that slot remains unchanged. Therefore, the induced EMF generated by the rotor acting on a particular stator slot after a demagnetization fault is distorted within one mechanical cycle.

[0004] The literature [Chen, H., Gao, CX, Si, JK, Nie, YJ, & Hu, YH (2021). A Novel Method for Diagnosing Demagnetization Fault in PMSM Using Toroidal-Yoke-Type Search Coil. IEEE Transactions on Instrumentation and Measurement, 71, 1–12] proposes installing a ring yoke coil on the stator side to obtain signals containing demagnetization fault information. Using the induced electromotive force of the ring yoke coil, a demagnetization location signal is constructed and partitioned according to certain rules. The waveform of the partitioned demagnetization location signal is used to identify the type of demagnetization fault and determine the state of the demagnetized permanent magnet in the corresponding area. This method achieves comprehensive coverage of different magnetic pole positions by arranging multiple ring yoke coils on the stator side of the motor, thus acquiring rich induced signals containing demagnetization fault information. However, this method is an invasive detection method, requiring complex detection devices and a large number of coils, thus limiting its practicality in engineering applications.

[0005] The literature [Song, JC, Zhao, JW, Dong, F., Zhao, J., Xu, L., & Yao, Z. (2020). A New Demagnetization Fault Recognition and Classification Method for DPMSLM. IEEE Transactions on Industrial Informatics, 16(3), 1559-1570] proposes a demagnetization fault diagnosis method based on air gap magnetic field measurement. This method involves arranging a gaussmeter within the motor's air gap to acquire a magnetic field signal containing the position information of the demagnetizing permanent magnet. A demagnetization fault analysis model is established using the finite element method, and the air gap magnetic flux density is extracted and transformed to construct corresponding fault features. Subsequently, machine learning algorithms are used to identify and locate the demagnetization position. This method enables direct measurement of the air gap magnetic field using a gaussmeter, providing intuitive feature information, but it increases the complexity of the motor structure and the difficulty of installation. Summary of the Invention

[0006] In view of the above, the present invention provides a method for locating demagnetization faults in permanent magnet synchronous motors based on branch current. The method is simple in principle, easy to implement, reduces the cost of diagnosis, and avoids the impact of installing unnecessary sensors on the normal operation of the motor.

[0007] A method for locating demagnetization faults in a permanent magnet synchronous motor based on branch current includes the following steps: (1) The three-phase stator current of the motor under test under different magnetic pole demagnetization faults was obtained by simulation. i a ~ i c and branch current i branch ; (2) Collect the actual three-phase stator current of the motor under test during operation. i a-real ~ i c-real ; (3) Construct multiple sets of simulation samples, each corresponding to each magnetic pole, which includes the three-phase stator current obtained by simulation under the demagnetization fault of that magnetic pole. i a ~ i c and branch current i branch ; (4) Construct a branch current prediction model based on GAN (Generative Adversarial Network) and train it using simulation samples; (5) The actual three-phase stator current i a-real ~ i c-real The input is fed into the trained branch current prediction model to predict the corresponding branch current. i* branch ; (6) The branch current i* branch Correlation calculations are performed with each simulated sample, and the result is taken as... i* branch The simulation sample with the highest correlation is used to determine the magnetic pole corresponding to the simulation sample, which is the location of the demagnetization fault.

[0008] Furthermore, the specific implementation of step (1) is as follows: First, based on the design parameters of the motor under test, a finite element model of the motor under test is established. In the finite element model, the three-phase stator current of the motor under test is simulated when different magnetic poles experience the same degree of demagnetization fault at rated speed. i a ~ i c and branch current i branch And record and save.

[0009] Furthermore, the branch current i branchBy setting up an external circuit in the finite element model, i.e., setting two parallel branches for each phase, involving winding connections and load impedance matching, the current in the first branch of the A-phase winding is obtained through simulation as the branch current. i branch .

[0010] Furthermore, in step (2), the actual three-phase stator current of the motor under test is collected by a current sensor during operation. i a-real ~ i c-real The sampling frequency is not less than 10kHz, and the three-phase current data is collected by using a data acquisition card through synchronous sampling technology.

[0011] Furthermore, the branch current prediction model in step (4) includes: Generator to simulate three-phase stator currents in the sample i a ~ i c As input, branch currents are generated through learning and prediction. i' branch ; The discriminator is based on the branch currents in the simulation sample. i branch Branch current of generator output i' branch Perform real / fake detection and optimize generator parameters through adversarial training.

[0012] Furthermore, in step (4), during the training of the branch current prediction model, MSE (mean squared error) is used as the loss function of the generator, and cross-entropy is used as the loss function of the discriminator. The training termination condition is the branch current. i' branch and i branch The correlation coefficient is greater than or equal to 0.95.

[0013] Furthermore, in step (5), the actual three-phase stator current is... i a-real ~ i c-real The input is fed into the generator of the trained branch current prediction model to predict and generate the corresponding branch current. i* branch .

[0014] Furthermore, before performing correlation calculations in step (6), the branch currents need to be analyzed. i* branch and the branch current in each simulation sample i branch Normalization is performed.

[0015] Furthermore, the correlation calculation in step (6) uses the Pearson correlation coefficient, which is obtained by calculating the branch current. i* branch With the branch current in each simulation sample i branch The Pearson correlation coefficient between the samples is used to determine the location of the demagnetization fault. The magnetic pole corresponding to the sample with the largest Pearson correlation coefficient is taken.

[0016] A computer device includes a memory and a processor, wherein the memory stores a computer program and the processor executes the computer program to implement the above-described method for locating demagnetization faults in a permanent magnet synchronous motor based on branch current.

[0017] A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described method for locating demagnetization faults in a permanent magnet synchronous motor based on branch current.

[0018] This invention is based on the principle of local demagnetization fault magnetic pole location using current signals combined with intelligent algorithms. It establishes a model based on current simulation data to express the mapping relationship from the three-phase current of the motor to the branch current, achieving accurate training of the mapping model with limited data. Therefore, this invention has the following beneficial technical effects: 1. The present invention provides a demagnetization fault diagnosis method based on current signal combined with intelligent algorithm. The method obtains model data and real-time acquired original motor signals through acquisition module and finite element model, then obtains the generated branch current signal through generative adversarial network, and finally determines which magnetic pole has demagnetized fault through Pearson correlation coefficient.

[0019] 2. Compared with traditional methods, the present invention avoids dependence on the sensor installation location and avoids the impact of sensor installation and search coil on the normal operation of the motor, and the diagnostic model is simple.

[0020] 3. The demagnetization fault location method of the present invention can locate and diagnose the demagnetized permanent magnet after partial demagnetization of the motor, so as to ensure the safe operation of the motor and reduce the property and safety problems caused by partial demagnetization faults. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the demagnetization fault location method for permanent magnet synchronous motors based on branch current according to the present invention.

[0022] Figure 2 This is a schematic diagram illustrating the structural principle of the generative adversarial network in this invention.

[0023] Figure 3The following is a schematic diagram of the branch current under different magnetic pole demagnetization faults, taking a 10-pole 12-slot motor as an example in the embodiment of the present invention. (a) to (h) correspond to the comparison results of simulated branch current and predicted branch current when demagnetization faults occur in 8 different magnetic poles.

[0024] Figure 4 This is a schematic diagram of the external circuit of the motor winding in this invention. Detailed Implementation

[0025] To describe the present invention in more detail, the technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0026] like Figure 1 As shown, the demagnetization fault location method for permanent magnet synchronous motors based on branch current of the present invention includes the following steps: (1) Based on the design parameters of the motor under test, establish the finite element model of the motor under test, and set up the external circuit to collect the simulated branch current. i branch .

[0027] The stator three-phase current of the motor under test when different permanent magnet poles experience the same degree of demagnetization fault at rated speed. i a , i b and i c and the current in the first branch of phase A winding i branch Record and save. The external circuitry includes... Figure 4 As shown, each phase is configured with two parallel branches, involving winding connection methods and load impedance matching. The current in the first branch of phase A winding is obtained through simulation. i branch1 As branch current i branch Right now i branch = i branch1 .

[0028] (2) The actual stator three-phase current of the motor under test during operation is collected by a current sensor. i a-real , i b-real and i c-real The sampling frequency is not less than 10kHz, and the three-phase current data is collected by using a data acquisition card through synchronous sampling technology.

[0029] (3) The stator three-phase currents corresponding to different magnetic pole demagnetization faults in the simulation model are used to calculate the currents. ia , i b and i c and branch current i branch As training samples, a set of stator three-phase currents and branch currents are used for each type of magnetic pole demagnetization case, thus obtaining multiple sets of training samples.

[0030] (4) Constructing such Figure 2 The generative adversarial network shown includes: Generator: Based on stator three-phase current i a , i b and i c As input, output predicted branch current i' branch ; Discriminator: Receives the actual branch current i branch With the generated i' branch The generator parameters are optimized through adversarial training.

[0031] (5) The branch current obtained from the finite element model i branch The data is fed into the discriminator as real data, and the corresponding three-phase current under this fault condition is fed into the generator as input data, and the generator is continuously trained in the generative adversarial network.

[0032] During training, the generator's loss function is MSE (mean squared error), the discriminator's loss function is cross-entropy, and the training termination condition is the generator's output. i' branch With reality i branch The correlation coefficient is greater than or equal to 0.95.

[0033] (6) Export the trained generator parameters and the signals acquired by the signal acquisition module. i a-real , i b-real and i c-real As input to the generator, the generated branch current is obtained through the generator.

[0034] (7) The first branch current in phase A winding of the motor under test when different permanent magnet poles of the motor under test experience the same degree of demagnetization fault at rated speed. i branch1 The generator produces the branch current. i'branch The data is normalized using the following formula:

[0035] in: X norm These are the normalized branch current data. X These are the raw branch current data. X min It is the minimum value in this set of data. X max It is the maximum value in this set of data.

[0036] The first branch current of phase A winding under all fault conditions used for comparison i branch1 The normalized data is then placed into the sample database [branch]. j Then the generated branch current i' branch The data after normalization is used as a data group e .

[0037] (8) Set up the data group e With sample database [branch] j The Pearson correlation coefficient of the branch current is calculated using the following formula:

[0038] in: n This indicates the amount of data collected by the signal. k Indicates the first k Data points, k =1,2,……, n ; Represents the standardized residual value e The average value of the data This represents each group of data in the sample database. b ] j The average value; Pearson correlation coefficient r j This indicates the correlation between two sets of data. Therefore, the larger the correlation coefficient between the two sets of data, the higher the similarity. The permanent magnet with the highest similarity is the demagnetized position when the motor is actually used.

[0039] Finally, the sample with the highest correlation coefficient was determined to be the magnetic pole of the demagnetization fault location.

[0040] This implementation takes a 10-pole, 12-slot motor as an example. A generative adversarial network (GAN) is trained using simulation data of the motor. Real-time collected raw three-phase current data of the motor is fed into the trained GAN to generate corresponding branch currents. By comparing the generated branch currents with those in the database, it is determined which pole has a demagnetization fault, thus achieving fault location of the motor's faulty pole. The branch current comparison results obtained using the above method are as follows: Figure 3 As shown, it can be seen that the simulated branch current and the predicted branch current have the highest similarity when the magnetic pole 2 experiences a demagnetization fault.

[0041] The conclusion drawn from the above is that the permanent magnet fault diagnosis method based on branch current combined with intelligent algorithm is simple to apply and has high accuracy. It eliminates the need to install additional sensors or components for fault location diagnosis of permanent magnet synchronous motors, thus avoiding the influence of additional sensors or components on the motor during normal operation.

[0042] The above description of the embodiments is provided to enable those skilled in the art to understand and apply the present invention. Those skilled in the art can readily make various modifications to the above embodiments and apply the general principles described herein to other embodiments without creative effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made to the present invention by those skilled in the art based on the disclosure thereof should be within the scope of protection of the present invention.

Claims

1. A method for locating demagnetization fault of permanent magnet synchronous motor based on branch current, characterized in that, The method comprises the following steps: (1) Obtain the three-phase stator currents of the motor under different magnetic pole demagnetization faults through simulation i a i c and branch currents i branch ;​ (2) collecting actual three-phase stator currents of the motor under operation i a-real i c-real ​​ (3) Constructing multiple groups of simulation samples, each simulation sample corresponding to each magnetic pole, containing three-phase stator currents simulated under the demagnetization failure of the magnetic pole i a i c and branch currents i branch ;​ (4) constructing a branch current prediction model based on GAN, and training the model by using simulation samples; (5) the actual three-phase stator currents i a-real i c-real are input into the generator of the branch current prediction model trained, and the corresponding branch currents i* branch are predicted.​ (6) The branch current i* branch Correlation calculations are performed with each simulated sample, and the result is taken as... i* branch The simulation sample with the highest correlation is used to determine the magnetic pole corresponding to the simulation sample, which is the location of the demagnetization fault.

2. The method of claim 1, wherein the method is characterized by: The specific implementation of the step (1) is: firstly, a finite element model of the measured motor is established according to the design parameters of the measured motor, and the three-phase stator currents of the measured motor when different magnetic poles occur the same degree of demagnetization failure at the rated speed are simulated in the finite element model i a i c and branch currents i branch are recorded and saved.​ 3. The method of claim 2, wherein the method is characterized by: The branch current i branch By setting the external circuit in the finite element model, that is, setting 2 parallel branches for each phase, involving winding connection and load impedance matching, and then simulating to obtain the current on the first branch in the A-phase winding as the branch current i branch .

4. The method of claim 1, wherein the method is characterized by: The step (2) collects the actual three-phase stator current of the measured motor in operation by the current sensor i a-real i c-real The sampling frequency is not less than 10 kHz, and the three-phase current data are collected by the synchronous sampling technology through the collection card.​ 5. The method of claim 1, wherein: The branch current prediction model in the step (4) comprises: Generator to emulate three-phase stator currents in a sample i a i c As input, predict generation of branch currents by learning i' branch ;​ discriminator, according to branch currents in the simulation sample i branch branch currents output by the generator i' branch true or false discrimination, optimization of generator parameters through adversarial training.

6. The method of claim 5, wherein: The step (4) adopts MSE as the loss function of the generator and adopts cross entropy as the loss function of the discriminator in the branch current prediction model training process, and the training termination condition is that the branch current i' branch The correlation coefficient of the i branch is greater than or equal to 0.

95.

7. The method of claim 1, wherein: The step (6) needs to normalize the branch current i* branch before the correlation calculation is performed i branch performed.

8. The method of claim 1, wherein: The correlation calculation in the step (6) adopts a Pearson correlation coefficient, and the Pearson correlation coefficient is calculated by calculating the Pearson correlation coefficient between the branch current i* branch of each simulation sample i branch The simulation sample corresponding to the maximum Pearson correlation coefficient is taken, and the magnetic pole corresponding to the simulation sample is the demagnetization fault position.

9. A computer device comprising a memory and a processor, said memory having stored therein a computer program, characterized in that: The processor is configured to execute the computer program to implement the method for locating demagnetization fault of the permanent magnet synchronous motor based on branch current according to any one of claims 1-8.

10. A computer readable storage medium storing a computer program, characterized in that: The computer program is configured to be executed by the processor to implement the method for locating demagnetization fault of the permanent magnet synchronous motor based on branch current according to any one of claims 1-8.

Images