CN-121721488-B - Motor demagnetization detection method, system and device based on VMD and DANN

CN121721488BCN 121721488 BCN121721488 BCN 121721488BCN-121721488-B

Abstract

The invention relates to the technical field of motor fault diagnosis, in particular to a motor demagnetization detection method, a motor demagnetization detection system and a motor demagnetization detection device based on VMD and DANN. The method comprises the steps of establishing current models of single magnetic poles of a tested motor under different demagnetizing degrees, constructing a plurality of groups of simulation samples, decomposing three-phase stator currents through VMDs to obtain feature vectors, constructing a DANN-based demagnetizing degree prediction model, training to obtain a trained demagnetizing degree prediction model, collecting actual three-phase stator currents of the tested motor during operation, performing VMD decomposition on the actual three-phase stator currents, inputting the actual three-phase stator currents into the DANN model, and predicting the demagnetizing degree of the actual motor. The invention utilizes the finite element joint simulation to generate the demagnetized sample data with the label, overcomes the limit that a large number of samples with known demagnetized degree are difficult to obtain in practical application, and realizes accurate quantitative prediction of the actual motor demagnetized state.

Inventors

WANG ZHIQIANG
YAN SHIHAO
Yan Xiayu
Yang Chaodie
ZHANG GUOZHENG

Assignees

天津工业大学

Dates

Publication Date: 20260508
Application Date: 20260226

Claims (9)

1. The motor demagnetization detection method based on the VMD and the DANN is characterized by comprising the following steps of: s1, establishing current models of single magnetic poles of a tested motor under different demagnetizing degrees; S2, constructing a plurality of groups of simulation samples according to the current model, wherein each group of simulation samples corresponds to one demagnetization degree, and acquiring three-phase stator currents under different demagnetization degrees; S3, decomposing the three-phase stator current through VMD, and combining non-integer harmonic eigenvector functions containing demagnetizing fault characteristics to obtain eigenvectors; S4, constructing a DANN-based demagnetizing degree prediction model, and training the demagnetizing degree prediction model by using a simulation sample and a feature vector to obtain a trained demagnetizing degree prediction model; S5, collecting actual three-phase stator current when the tested motor runs, performing VMD decomposition on the actual three-phase stator current, inputting the feature vector obtained by decomposition into a trained demagnetization degree prediction model, and predicting the demagnetization degree of the actual motor.
2. The motor demagnetization detection method based on the VMD and the DANN of claim 1 is characterized in that the current model is a joint simulation model built based on Simulink, maxwell and Simplorer and comprises three parts, namely a SVPWM control algorithm model based on the SVPWM control algorithm in Simulink, a motor finite element simulation model in Maxwell and an inverter circuit model in Simplorer.
3. The method for detecting motor demagnetization based on VMD and DANN according to claim 1, wherein step S2 comprises the steps of simulating different demagnetization degrees by modifying residual magnetic induction intensity of a permanent magnet, setting single magnetic pole demagnetization degrees to be 100%, 75%, 50% and 25%, obtaining current data of different demagnetization degrees from simulation results, and forming training samples by each group of current data and corresponding demagnetization degree labels.
4. The method for detecting motor demagnetization based on VMD and DANN according to claim 1, wherein in step S3, a VMD algorithm is adopted to decompose three-phase stator current signals to obtain a series of eigen mode functions, and eigen mode functions including non-integer harmonic components of demagnetization fault characteristics are combined to be used as eigenvectors for diagnosing demagnetization degree.
5. The method for detecting motor demagnetization based on VMD and DANN according to claim 4, wherein the VMD algorithm is: S11, carrying out Hilbert transformation on the phase stator current signals to obtain a single-side frequency spectrum of each IMF; S12, combining the unilateral frequency spectrum with the estimated center frequency, and modulating the L2 norm of the signal gradient to obtain the bandwidth corresponding to each mode; and S13, constructing an augmented Lagrangian function according to the bandwidth, and solving to obtain a feature vector.
6. The method for detecting motor demagnetization based on VMD and DANN according to claim 1, wherein the DANN-based demagnetization degree prediction model includes the following modules: The feature extractor is used for taking the feature vector constructed after the VMD decomposes the stator current as input and extracting depth features related to the demagnetizing degree; The tag predictor is used for taking the information output by the feature extractor as input, predicting the demagnetization degree and outputting corresponding demagnetization state information; the domain discriminator is used for discriminating whether the input is from a simulation domain or an actual domain, and the feature extractor is forced to generate domain-invariant fault features through countermeasure training of the gradient inversion layer.
7. The method for detecting motor demagnetization based on VMD and DANN according to claim 6, wherein step S5 is that actual three-phase stator current is collected when the motor to be detected runs, after feature vectors are processed and constructed through VMD algorithm, the feature extractor and the label predictor in the model are transmitted forwards, and the demagnetization degree prediction result of the actual current feature vectors is output.
8. A VMD and DANN-based motor demagnetization detection system for performing a VMD and DANN-based motor demagnetization detection method according to any of claims 1 to 7, comprising: The model building module is used for building current models of single magnetic poles of the tested motor under different demagnetizing degrees; the current generation module is used for constructing a plurality of groups of simulation samples according to the current model, wherein each group of simulation samples corresponds to one demagnetization degree, and three-phase stator currents under different demagnetization degrees are obtained; the characteristic vector generation module is used for decomposing the three-phase stator current through the VMD, and combining non-integer harmonic eigenvector functions containing demagnetizing fault characteristics to obtain a characteristic vector; The model training module is used for constructing a demagnetizing degree prediction model based on the DANN, and training the demagnetizing degree prediction model by using the simulation sample and the feature vector to obtain a trained demagnetizing degree prediction model; the actual result prediction module is used for collecting actual three-phase stator current when the tested motor runs, performing VMD decomposition on the actual three-phase stator current, inputting the feature vector obtained by decomposition into the DANN model, and predicting the demagnetization degree of the actual motor.
9. An electronic device comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, when executing a computer program, implements the steps of a VMD and DANN-based motor demagnetization detection method according to any of claims 1 to 7.

Description

Motor demagnetization detection method, system and device based on VMD and DANN Technical Field The invention relates to the technical field of motor fault diagnosis, in particular to a motor demagnetization detection method, a motor demagnetization detection system and a motor demagnetization detection device based on VMD and DANN. Background Permanent magnet synchronous motors have been widely used in the fields of new energy automobiles, rail transit, aerospace, industrial transmission and the like because of high efficiency, high power density and excellent control performance. However, in the long-term operation process, the permanent magnet may be locally demagnetized or irreversibly demagnetized due to factors such as high temperature, overcurrent, short-circuit impact, mechanical stress, etc., so that the distribution distortion of the air-gap field of the motor and the amplitude of back electromotive force are reduced, and non-integer multiple harmonic components are introduced into the stator current, thereby reducing the motor performance and even causing system failure. The method for accurately detecting the demagnetization faults and the degree thereof is beneficial to guaranteeing the operation safety of the motor, reducing the energy consumption, prolonging the service life, reflecting the development stage and the severity of the demagnetization, providing graded management and refined decision support for operation and maintenance personnel, allowing the motor to continue to operate and reasonably arrange maintenance plans when the light demagnetization occurs, adopting protection measures such as derating operation when the light demagnetization occurs, and timely stopping and overhauling when the demagnetization reaches the severity, and avoiding further expansion of the faults or causing system failure. Therefore, the demagnetization degree detection has important engineering significance for realizing hierarchical management, risk assessment and full life cycle health monitoring of the motor. Disclosure of Invention The present invention is directed to solving at least one of the technical problems existing in the related art. Therefore, the invention provides a motor demagnetization detection method, a motor demagnetization detection system and a motor demagnetization detection device based on VMD and DANN, which overcome the limitation that a large number of samples with known demagnetization degree are difficult to obtain in practical application, and realize accurate quantitative prediction of the actual motor demagnetization state. The invention provides a motor demagnetization detection method based on VMD and DANN, which comprises the following steps: s1, establishing current models of single magnetic poles of a tested motor under different demagnetizing degrees; S2, constructing a plurality of groups of simulation samples according to the current model, wherein each group of simulation samples corresponds to one demagnetization degree, and acquiring three-phase stator currents under different demagnetization degrees; S3, decomposing the three-phase stator current through VMD, and combining non-integer harmonic eigenvector functions containing demagnetizing fault characteristics to obtain eigenvectors; S4, constructing a DANN-based demagnetizing degree prediction model, and training the demagnetizing degree prediction model by using a simulation sample and a feature vector to obtain a trained demagnetizing degree prediction model; S5, collecting actual three-phase stator current when the tested motor runs, performing VMD decomposition on the actual three-phase stator current, inputting the feature vector obtained by decomposition into a trained demagnetization degree prediction model, and predicting the demagnetization degree of the actual motor. The motor demagnetization detection method based on the VMD and the DANN provided by the invention is characterized in that the current model is a joint simulation model built based on Simulink, maxwell and Simplorer, and comprises three parts, namely a SVPWM control algorithm model based in Simulink, a motor finite element simulation model in Maxwell and an inverter circuit model in Simplorer. According to the motor demagnetization detection method based on the VMD and the DANN, the step S2 comprises the steps of simulating different demagnetization degrees by modifying residual magnetic induction intensity of a permanent magnet, setting the demagnetization degree of a single magnetic pole to be 100%, 75%, 50% and 25%, obtaining current data of different demagnetization degrees from simulation results, and forming a training sample by each group of current data and corresponding demagnetization degree labels. According to the motor demagnetization detection method based on VMD and DANN, in step S3, a VMD algorithm is adopted to decompose three-phase stator current signals to obtain a series of eigen mode functions, and the eigen mode f