Motor rotor anomaly detection method, device and equipment

Abstract

The invention relates to a motor rotor anomaly detection method, device and equipment. The method comprises the following steps: acquiring current vibration data of a motor rotor at a current moment and historical vibration data of the motor rotor at a historical moment; performing feature enhancement on similar features in the current vibration data to obtain first enhanced features; performing feature enhancement on the first enhancement feature according to the current vibration data and the historical vibration data to obtain a second enhancement feature; performing data reconstruction on the second enhanced feature to obtain reconstructed vibration data; and determining whether the current vibration data is abnormal or not according to the reconstructed vibration data and the current vibration data. According to the invention, anomaly detection can be carried out on the motor rotor in real time, off-line inspection is not needed, and needed manpower and material resources are reduced.

Description

Technical Field

[0001] The present application relates to the field of motor technology, and in particular to a method, device and equipment for detecting motor rotor abnormality. Background Art

[0002] Electric vehicles, as the future direction of transportation, are seeing their market share increase annually. The motor is a core component of electric vehicles, and its performance and reliability are directly related to the vehicle's operating efficiency and safety. However, due to long-term operation in complex operating environments, motors are prone to various faults. As a crucial component of the motor, rotor anomalies can significantly impact its performance and operation. Rotor anomalies primarily manifest as broken rotor bars and rotor imbalance. Broken rotor bars can lead to unstable motor operation, abnormal vibration and noise, and in severe cases, mechanical damage. Rotor imbalance can be caused by manufacturing errors, wear from long-term operation, or improper maintenance, impacting the motor's stability and service life. Rotor anomalies not only degrade electric vehicle performance but can also cause serious safety incidents. Of course, motor rotors are used not only in electric vehicles but also in other equipment, where these anomalies can also occur.

[0003] To this end, rotor anomaly detection is necessary. Related technologies include regular vibration analysis and mechanical testing of the rotor to ensure optimal rotor operation, extend its service life, and reduce the probability of failure. However, regular vibration analysis and mechanical testing require offline inspection of the motor rotor, which consumes a lot of manpower and material resources. Summary of the Invention

[0004] Based on this, it is necessary to provide a motor rotor abnormality detection method, device and equipment that can detect abnormalities of the motor rotor in real time to address the above technical problems.

[0005] In a first aspect, the present application provides a method for detecting an abnormality of a motor rotor, comprising:

[0006] Obtain the current vibration data of the motor rotor at the current moment and the historical vibration data at the historical moments;

[0007] Performing feature enhancement on similar features in the current vibration data to obtain a first enhanced feature;

[0008] Performing feature enhancement on the first enhanced feature according to the current vibration data and the historical vibration data to obtain a second enhanced feature;

[0009] reconstructing data of the second enhanced feature to obtain reconstructed vibration data;

[0010] According to the reconstructed vibration data and the current vibration data, it is determined whether the current vibration data is abnormal.

[0011] In one embodiment, similar features in the current vibration data are enhanced to obtain first enhanced features, including: determining a first enhancement factor according to the current vibration data; the first enhancement factor representing the similarity between the vibration signals in the current vibration data; performing feature enhancement on the current vibration data according to the first enhancement factor to obtain initial enhanced features; determining a second enhancement factor according to the initial enhanced features; the second enhancement factor representing the similarity between the features in the initial enhanced features; performing feature enhancement on the initial enhanced features according to the second enhancement factor to obtain the first enhanced features.

[0012] In one embodiment, the first enhancement factor is determined according to the current vibration data, including: using at least two similarity calculation methods to determine the first similarity degree data between different vibration signals in the current vibration data; determining the first enhancement factor according to the first similarity degree data.

[0013] In one embodiment, the first enhancement factor is determined according to the first similarity degree data, including: determining a data fusion coefficient according to the first similarity degree data; performing data fusion on the first similarity degree data according to the data fusion coefficient to obtain degree fusion data; performing dimension reduction processing on the degree fusion data to obtain the first enhancement factor.

[0014] In one embodiment, the first enhanced features are enhanced according to the current vibration data and the historical vibration data to obtain second enhanced features, including: obtaining third enhanced features corresponding to the historical vibration data; the third enhanced features being the result of feature enhancement on similar features in the historical vibration data; determining a third enhancement factor according to the first enhanced features and the third enhanced features; the third enhancement factor representing the feature similarity between the first enhanced features and the third enhanced features; performing feature enhancement on the first enhanced features according to the third enhancement factor to obtain the second enhanced features.

[0015] In one embodiment, the third enhancement factor is determined according to the first enhanced features and the third enhanced features, including: using a preset similarity calculation method to determine second similarity degree data between the features in the first enhanced features; obtaining third similarity degree data corresponding to the third enhanced features; the third similarity degree data representing the similarity between the features in the third enhanced features; determining the third enhancement factor according to the second similarity degree data and the third similarity degree data.

[0016] In one embodiment, the first enhancement feature is feature enhanced according to the third enhancement factor to obtain the second enhancement feature, including: determining a fused vibration feature according to the first enhancement feature and the third enhancement feature; and feature enhancing the first enhancement feature according to the fused vibration feature and the third enhancement factor to obtain the second enhancement feature.

[0017] In one embodiment, determining whether the current vibration data is abnormal based on the reconstructed vibration data and the current vibration data includes: determining a difference between the reconstructed vibration data and the current vibration data; and determining whether the current vibration data is abnormal based on the difference.

[0018] In a second aspect, the present application further provides a motor rotor abnormality detection device, comprising:

[0019] A data acquisition module is used to acquire the current vibration data of the motor rotor at the current moment and the historical vibration data at the historical moments;

[0020] A first enhancement module is used to enhance similar features in the current vibration data to obtain a first enhanced feature;

[0021] A second enhancement module is used to enhance the first enhancement feature according to the current vibration data and the historical vibration data to obtain a second enhancement feature;

[0022] A data reconstruction module, configured to reconstruct data of the second enhanced feature to obtain reconstructed vibration data;

[0023] The abnormality determination module is used to determine whether the current vibration data has an abnormality based on the reconstructed vibration data and the current vibration data.

[0024] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the method provided in the first aspect when executing the computer program.

[0025] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the method provided in the first aspect when the computer program is executed by a processor.

[0026] In a fifth aspect, the present application also provides a computer program product, comprising a computer program, which implements the method provided in the first aspect when executed by a processor.

[0027] The above-mentioned motor rotor abnormality detection method, device and equipment perform two feature enhancements on the current vibration data, thereby strengthening important features, effectively reducing the influence of noise introduction and redundant features, and helping to improve the accuracy of subsequent abnormality detection. Then, based on the current vibration data and the reconstructed vibration data obtained after data reconstruction of the second enhanced feature, it is determined whether the current vibration data has an abnormality. It can be seen that the abnormality detection process is relatively simple and easy to implement. Since the present application can detect whether the motor rotor has an abnormality in real time, the abnormality of the motor rotor can be discovered in time, thereby performing maintenance warnings, improving the service life of the motor rotor, and improving the reliability and safety of the motor operation. Moreover, since no offline inspection is required, the motor operation efficiency can be improved, the required manpower and material resources can be reduced, and the personnel do not need to have strong professionalism, thereby reducing maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments of the present application or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0029] Figure 1 1 is a flow chart of a method for detecting abnormality of a motor rotor in one embodiment;

[0030] Figure 2 FIG1 is a flow chart of the steps of obtaining the first enhancement feature in one embodiment;

[0031] Figure 3 FIG1 is a flow chart of the step of determining the first enhancement factor in one embodiment;

[0032] Figure 4 FIG1 is a flow chart of the step of determining the first enhancement factor in one embodiment;

[0033] Figure 5 FIG1 is a flow chart of the steps of obtaining the second enhanced feature in one embodiment;

[0034] Figure 6 FIG1 is a flow chart of the step of determining the third enhancement factor in one embodiment;

[0035] Figure 7 A schematic flow chart of a feature enhancement step of a first enhancement feature in one embodiment;

[0036] Figure 8 Schematic diagram of a flow chart of an abnormality determination step in one embodiment;

[0037] Figure 9A flowchart of the process of the obtaining step of the initial enhancement feature in one embodiment;

[0038] Figure 10 A flowchart of the process of the determining step of the third enhancement factor in one embodiment;

[0039] Figure 11 A structure block diagram of the motor rotor abnormality detection device in one embodiment;

[0040] Figure 12 An internal structure diagram of the computer device in one embodiment. DETAILED DESCRIPTION

[0041] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0042] In one exemplary embodiment, a motor rotor abnormality detection method is provided. Referring to Figure 1 , the motor rotor abnormality detection method comprises:

[0043] S110, obtaining current vibration data of the motor rotor at a current time and historical vibration data at a historical time.

[0044] Among the current vibration data and the historical vibration data, each includes a plurality of vibration signals, and the number of vibration signals in the current vibration data is the same as the number of vibration signals in the historical vibration data.

[0045] Among the current vibration data and the historical vibration data, each includes a plurality of vibration signals, and the number of vibration signals in the current vibration data is the same as the number of vibration signals in the historical vibration data.

[0046] Among the current vibration data and the historical vibration data, each includes a plurality of vibration signals, and the number of vibration signals in the current vibration data is the same as the number of vibration signals in the historical vibration data.

[0047] Among the current vibration data and the historical vibration data, each includes a plurality of vibration signals, and the number of vibration signals in the current vibration data is the same as the number of vibration signals in the historical vibration data.

[0048] In an optional implementation, an acceleration sensor is used to collect current vibration data of the motor rotor, thereby obtaining the current vibration data from the acceleration sensor.

[0049] S120 , performing feature enhancement on similar features in the current vibration data to obtain a first enhanced feature.

[0050] The first enhanced feature is obtained by enhancing the feature with higher similarity in the current vibration data and weakening the feature with lower similarity.

[0051] Understandably, the current vibration data collection process is inevitably subject to interference from various noise sources, such as environmental noise and electronic equipment noise. The vibration signal generated by the normal operation of the motor rotor exhibits a certain degree of periodicity and regularity, while noise is often random and irregular. Therefore, features with high similarity are considered important, while features with low similarity are considered noise. By enhancing similar features in the current vibration data, noise can be suppressed, important features can be highlighted, and data quality can be improved. Because noise signals are suppressed, the weak vibration signals generated by early faults are less likely to be overwhelmed by normal vibration signals, thereby improving the sensitivity of subsequent anomaly detection.

[0052] In actual scenarios, the process of feature enhancement in S120 involves a feature extraction process, which may include, but is not limited to, extracting features from current vibration data in a downsampling manner.

[0053] S130 , performing feature enhancement on the first enhanced feature according to the current vibration data and the historical vibration data to obtain a second enhanced feature.

[0054] The second enhanced feature is the result of enhancing the features in the first enhanced feature that have a high similarity to the historical vibration data, and weakening the features in the first enhanced feature that have a low similarity to the historical vibration data. That is, the feature enhancement of the first enhanced feature is performed based on the feature similarity between each vibration signal in the current vibration data and the corresponding vibration signal in the historical vibration data. The features in the first enhanced feature that have a high similarity to the historical vibration data are enhanced, and the features in the first enhanced feature that have a low similarity to the historical vibration data are weakened, resulting in the second enhanced feature. This shows that the features in the current vibration data that have a high similarity to the historical vibration data are important features.

[0055] It is understandable that, since vibration data is temporally correlated, mining the strongly correlated features between current vibration data and historical vibration data helps capture the hidden temporal correlation in the current vibration data, thereby improving the characterization capability of the second enhanced feature.

[0056] S140 , reconstructing data of the second enhanced feature to obtain reconstructed vibration data.

[0057] The reconstructed vibration data is the result of restoring the vibration data of the second enhanced feature. Because the important features in the second enhanced feature are highlighted and noise is suppressed, the reconstructed vibration data obtained after data reconstruction has less noise and higher data quality than the current vibration data.

[0058] The data reconstruction process may refer to the feature extraction process. For example, if the feature extraction process is implemented by downsampling, the data reconstruction process may be implemented by upsampling.

[0059] For example, the process from the current vibration data to the second enhanced features undergoes two feature extractions, or downsampling, and the data reconstruction process uses a two-layer multilayer perceptron (MLP) decoder. The first MLP layer in the decoder upsamples the second enhanced features once to obtain the upsampled result, and the second MLP layer in the decoder upsamples the upsampled result once to obtain the reconstructed vibration data. The MLP can be either a non-shared parameter MLP or a shared parameter MLP.

[0060] S150 , determining whether the current vibration data is abnormal based on the reconstructed vibration data and the current vibration data.

[0061] It can be understood that based on the feature similarities between the vibration signals in the current vibration data, similar features in the current vibration data are enhanced, achieving the first feature enhancement. Then, based on the feature similarities between the vibration signals in the current vibration data and the corresponding vibration signals in the historical vibration data, the first enhanced features are enhanced, achieving the second feature enhancement. Through these two feature enhancements, important features are enhanced, that is, important features are strengthened and non-important features are weakened, effectively reducing the impact of noise introduction and redundant features, and helping to improve the accuracy of anomaly detection.

[0062] It is understandable that, if the current vibration data does not contain an abnormal vibration signal, even after two feature enhancements and data reconstructions, the difference between the reconstructed vibration data and the current vibration data is mainly caused by noise, and the difference is relatively small. However, if the current vibration data contains an abnormal vibration signal, after two feature enhancements and data reconstructions, the difference between the reconstructed vibration data and the current vibration data increases due to the presence of the abnormal vibration signal. Therefore, the degree of difference between the reconstructed vibration data and the current vibration data can be used to determine whether the current vibration data is abnormal.

[0063] It can be seen that in the above-mentioned motor rotor abnormality detection method, the current vibration data is subjected to two feature enhancements, which realizes the strengthening of important features, effectively reduces the influence of noise introduction and redundant features, and helps to improve the accuracy of subsequent abnormality detection. Then, based on the current vibration data and the reconstructed vibration data obtained after data reconstruction of the second enhanced feature, it is determined whether the current vibration data has an abnormality. It can be seen that the abnormality detection process is relatively simple and easy to implement. Since the present application can detect whether the motor rotor has an abnormality in real time, the abnormality of the motor rotor can be discovered in time, thereby performing maintenance warnings, improving the service life of the motor rotor, and improving the reliability and safety of the motor operation. Moreover, since no offline inspection is required, the motor operation efficiency can be improved, the required manpower and material resources can be reduced, and the personnel do not need to have strong professionalism, thereby reducing maintenance costs.

[0064] Based on the technical solutions of the above embodiments, an optional embodiment is provided. In this optional embodiment, the step of obtaining the first enhanced feature in S120 is refined.

[0065] See also Figure 2 , the step of obtaining the first enhanced feature includes:

[0066] S210: Determine a first enhancement factor according to current vibration data.

[0067] The first enhancement factor represents the similarity between the vibration signals in the current vibration data.

[0068] For example, the current vibration data includes N vibration signals, and the first enhancement factor is a first array including N elements, where each element in the first array corresponds one-to-one to each vibration signal in the current vibration data, and each element in the first array represents the overall similarity between the corresponding vibration signal in the current vibration data and each other vibration signal. N is an integer greater than 2.

[0069] S220 , performing feature enhancement on the current vibration data according to the first enhancement factor to obtain an initial enhanced feature.

[0070] The initial enhanced features are obtained by enhancing the features with higher similarity in the current vibration data and weakening the features with lower similarity. S220 is the first feature enhancement of the current vibration data.

[0071] In an optional implementation, S220 may include: performing data enhancement on the current vibration data using a first enhancement factor to obtain enhanced vibration data; and performing feature extraction on the enhanced vibration data to obtain initial enhanced features.

[0072] It is understood that if the current vibration data includes N vibration signals, the enhanced vibration data includes N enhanced vibration signals, and the initial enhanced vibration signature includes N features. There is a one-to-one correspondence between the vibration signals in the current vibration data, the enhanced vibration signals in the enhanced vibration data, and the features in the initial enhanced vibration signature.

[0073] Exemplarily, each element in the first enhancement factor is multiplied by the corresponding vibration signal in the current vibration data to obtain a corresponding enhanced vibration signal, and each enhanced vibration signal forms enhanced vibration data.

[0074] The enhanced vibration data can be input into the first-layer shared parameter MLP in the VAE encoder (Variational Auto Encoder). The first-layer shared parameter MLP performs feature extraction to obtain the initial enhanced vibration features. In other words, the first-layer shared parameter MLP performs a downsampling of the enhanced vibration data.

[0075] It can be understood that shared parameter MLP means that the same MLP is used to extract features from vibration data at different times. The parameter sharing between each MLP can reduce the training time of the MLP.

[0076] S230: Determine a second enhancement factor according to the initial enhancement feature.

[0077] The second enhancement factor represents the similarity between the features in the initial enhancement feature.

[0078] The process of determining the second enhancement factor is similar to the process of determining the first enhancement factor, and thus reference may be made to S310 and S320 below to determine the second enhancement factor.

[0079] S240: Perform feature enhancement on the initial enhanced feature according to the second enhancement factor to obtain a first enhanced feature.

[0080] The first enhanced features are obtained by enhancing the features with higher similarity among the initial enhanced features and weakening the features with lower similarity. S240 is equivalent to the second feature enhancement of the current vibration data. Because the initial enhanced features may include some features that are valuable for subsequent analysis but not prominent enough, the second feature enhancement amplifies these valuable features.

[0081] Exemplarily, each element in the second enhancement factor is multiplied by the corresponding feature in the initial enhancement feature to obtain the first enhancement feature.

[0082] Exemplarily, the initial enhanced features include N features, the second enhancement factor is a second array including N elements, the first enhanced features include N features, each element in the second array, each feature in the initial enhanced features, and each feature in the first enhanced features have a one-to-one correspondence, and each element in the second array represents the overall similarity between the corresponding feature in the initial enhanced features and each other feature.

[0083] In practical scenarios, since the VAE encoder includes a second-layer shared parameter MLP in addition to the first-layer shared parameter MLP, after obtaining the first enhanced features, the first enhanced features can also be input into the second-layer shared parameter MLP. The second-layer shared parameter MLP then performs dimensionality reduction, i.e., downsampling, on the first enhanced features to obtain the reduced-dimensional first enhanced features. This shows that the second-layer shared parameter MLP compresses and abstracts the first enhanced features, capturing their important features.

[0084] During the neural network training process of the VAE encoder, the second-layer shared parameter MLP can be trained and optimized through divergence and loss functions, thereby improving the ability of the second-layer shared parameter MLP to retain important features during the dimensionality reduction process.

[0085] As can be seen, first, the first enhancement factor is determined and used to enhance the current vibration data to obtain the initial enhanced feature. Then, the second enhancement factor is determined and used to enhance the initial enhanced feature to obtain the first enhanced feature. In other words, after two feature enhancements, the first enhanced feature is obtained, which can achieve a higher enhancement effect and help improve the accuracy of subsequent anomaly detection.

[0086] In actual scenarios, a single feature enhancement can be used to obtain the first enhanced feature, or three or more feature enhancements can be used to obtain the first enhanced feature. The more times the feature is enhanced, the more time is required. Therefore, excessive feature enhancement will affect the real-time requirements of anomaly detection. The enhancement effect of the first enhanced feature obtained by a single feature enhancement is slightly weaker than that obtained by the above-mentioned two feature enhancements. Therefore, the enhancement effect of the first enhanced feature obtained by two feature enhancements is better, ensuring both computational efficiency and accuracy. The enhancement effect refers to the degree to which important features are strengthened and the degree to which unimportant features are weakened.

[0087] Based on the technical solutions of the above embodiments, an optional embodiment is provided. In this optional embodiment, the step of determining the first enhancement factor in S210 is refined.

[0088] See also Figure 3 , the step of determining the first enhancement factor includes:

[0089] S310 , using at least two similarity calculation methods to respectively determine first similarity data between different vibration signals in the current vibration data.

[0090] The similarity calculation method can be selected as needed, for example, a Euclidean distance calculation method, a cosine similarity calculation method, etc. Of course, other similarity calculation methods can also be selected, which are not limited here.

[0091] The first similarity data reflects the similarity between different vibration signals in the current vibration data. The similarity between different vibration signals in the current vibration data can be accurately determined by similarity calculation.

[0092] In an optional implementation, the process of determining the first similarity data includes:

[0093] 1. Perform feature extraction on the current vibration data to obtain the current vibration feature vector. Each element in the current vibration feature vector corresponds one-to-one to each vibration signal in the current vibration data.

[0094] A non-shared parameter MLP can be used to extract features from the current vibration data. This means that different MLPs are used to extract features from vibration data at different times. The parameters of each MLP are not shared, thus avoiding parameter interference between feature extraction tasks. In other words, the non-shared parameter MLP is used to downsample the current vibration data.

[0095] 2. Multiply the current vibration eigenvector by the transpose of the current vibration eigenvector to obtain a first correlation matrix as a first similarity data.

[0096] For example, the current vibration data X t The current vibration eigenvector is M t , current vibration data X t It includes N vibration signals, and the current vibration feature vector M t The size is N*1, the current vibration feature vector M t The transpose of M t T , M t T The size is 1*N, M t With M t T Multiply them together to get the first incidence matrix S 1t , S 1t The size of is N*N. t is the current moment. The features in the current vibration eigenvector are dimensionless, so the elements in the first incidence matrix are dimensionless.

[0097] As can be seen, this implementation method obtains the first similarity data by multiplying the current vibration eigenvector by the transpose of the current vibration eigenvector. Since the first similarity data is calculated using the current vibration eigenvector rather than directly using the current vibration data, this is an indirect similarity calculation method.

[0098] In another optional implementation, the process of determining the first similarity data includes:

[0099] 1. Calculate the Euclidean distance between each vibration signal in the current vibration data to obtain a second correlation matrix as a first similarity data; wherein each row element in the second correlation matrix corresponds one-to-one to each vibration signal in the current vibration data, and each row element in the second correlation matrix is ​​the Euclidean distance between the corresponding vibration signal in the current vibration data and each other vibration signal.

[0100] For example, the Euclidean distance between the i-th vibration signal and the j-th vibration signal in the current vibration data is calculated using the following formula:

[0101]

[0102] Where, is the Euclidean distance between the i-th vibration signal and the j-th vibration signal in the current vibration data, is the i-th vibration signal in the current vibration data, is the jth vibration signal in the current vibration data. i and j are both integers greater than or equal to 1 and less than or equal to N. In practical scenarios, the calculated Euclidean distance is normalized to remove the dimension for easier calculation.

[0103] The above formula can be used to calculate the Euclidean distances between the i-th vibration signal in the current vibration data and the N vibration signals in the current vibration data, obtaining N Euclidean distances. These N Euclidean distances are used as the i-th row elements in the second incidence matrix. Using the above method, the row elements in the second incidence matrix can be obtained.

[0104] It can be seen that this implementation method determines the first similarity data by calculating the Euclidean distance between the vibration signals in the current vibration data. Since the first similarity data is calculated directly using the current vibration data, it is a direct similarity calculation method.

[0105] It can be seen that the above two implementations provide two similarity calculation methods. Of course, other similarity calculation methods can also be used, which are not limited here.

[0106] S320: Determine a first enhancement factor according to each piece of first similarity data.

[0107] It can be seen that the S310 and the S320 calculate a plurality of first similarity degree data through a plurality of similarity calculation manners, and determine the first enhancement factor by using the first similarity degree data. Since the plurality of similarity calculation manners can capture more angle similarity data, the effectiveness of the first enhancement factor can be improved, thereby improving the effect of feature enhancement of the current vibration data by using the first enhancement factor. Moreover, the first similarity degree data calculated by a single similarity calculation manner can be avoided from having deviation, thereby ensuring the accuracy of the first enhancement factor.

[0108] On the basis of the technical solutions of the above embodiments, an optional embodiment is further provided, in which the determination step of the first enhancement factor in the S320 is refined.

[0109] Referring to Figure 4 , the determination step of the first enhancement factor comprises:

[0110] S410, determining a data fusion coefficient according to the first similarity degree data.

[0111] The data fusion coefficient is data representing the correlation degree between the first similarity degree data, and can quantify the weight of the first similarity degree data in fusion, thereby fusing the first similarity degree data.

[0112] The data fusion coefficient comprises N coefficients, and the N coefficients correspond to the N vibration signals in the current vibration data one by one.

[0113] Exemplarily, the qth coefficient in the data fusion coefficient is determined by using the following calculation formula:

[0114]

[0115] In the formula, is the qth coefficient, is the qth row element of the qth vibration signal in the current vibration data in the second correlation matrix, is the qth row element of the qth vibration signal in the current vibration data in the first correlation matrix. q is an integer greater than or equal to 1 and less than N. Symbol represents a norm, for example, a Euclidean norm. The data fusion coefficient is dimensionless.

[0116] It can be seen that the above calculation formula calculates the data fusion coefficient by using the cosine similarity.

[0117] S420, fusing the first similarity degree data by using the data fusion coefficient, to obtain degree fusion data.

[0118] Among them, the degree fusion data integrates the information of each first similarity degree data, gathers the characteristics and advantages of different first similarity degree data, and reflects the similarity between different vibration signals in the current vibration data more comprehensively and accurately than a single first similarity degree data. That is, the degree fusion data reflects the comprehensive similarity between each vibration signal in the current vibration data.

[0119] The size of each first similarity data is N*N, and the size of the degree fusion data is N*N.

[0120] Exemplarily, the following calculation formula is used to fuse the qth row elements in each first similarity degree data to obtain the qth row elements in the degree fusion data:

[0121]

[0122] in, is the qth row element in the degree fusion data.

[0123] S430: Perform dimensionality reduction processing on the degree fusion data to obtain a first enhancement factor.

[0124] For example, the size of the degree fusion data is N*N, and the size of the first enhancement factor is N*1. The degree fusion data needs to be subjected to dimensionality reduction to obtain the first enhancement factor. The first enhancement factor is dimensionless. Of course, the second and third enhancement factors are also dimensionless.

[0125] Exemplarily, the following calculation formula is used to map the qth row element in the degree fusion data to the qth element in the first enhancement factor:

[0126]

[0127] Where, is the qth element in the first enhancement factor, is the activation function. The N elements in the first enhancement factor can be obtained through the above calculation formula, that is, the first enhancement factor .

[0128] As can be seen, by fusing the first similarity data using the data fusion coefficient, the resulting fused data reflects the comprehensive similarity between the vibration signals in the current vibration data. By performing dimensionality reduction on the fused data, a first enhancement factor of the desired size can be obtained, which can also represent the comprehensive similarity between the vibration signals in the current vibration data.

[0129] Based on the technical solutions of the above embodiments, an optional embodiment is provided. In this optional embodiment, the step of obtaining the second enhanced feature in S130 is refined.

[0130] See also Figure 5 , the step of obtaining the second enhanced feature includes:

[0131] S510: Acquire a third enhanced feature corresponding to historical vibration data.

[0132] The third enhanced feature is a result of enhancing similar features in the historical vibration data.

[0133] It is understandable that, at the previous moment, the third enhanced feature may be determined according to S210 to S240.

[0134] S520: Determine a third enhancement factor according to the first enhancement feature and the third enhancement feature.

[0135] The third enhancement factor represents the feature similarity between the first enhancement feature and the third enhancement feature.

[0136] It is understandable that, since the third enhancement factor represents the feature similarity between the first enhancement feature and the third enhancement feature, the third enhancement factor indirectly represents the feature similarity between the current vibration data and the historical vibration data.

[0137] S530: Perform feature enhancement on the first enhancement feature according to the third enhancement factor to obtain a second enhancement feature.

[0138] The second enhanced feature is obtained by enhancing the features of the first enhanced feature that are more similar to the historical vibration data and weakening the features of the first enhanced feature that are less similar to the historical vibration data. That is, by enhancing the first enhanced feature, the hidden time correlation in the current vibration data can be captured, and the time correlation representation ability of the second enhanced feature can be improved.

[0139] It can be seen that since the third enhancement factor indirectly represents the feature similarity between the current vibration data and the historical vibration data, using the third enhancement factor to perform feature enhancement on the first enhancement feature can accurately enhance the features in the first enhancement feature that are similar to the third enhancement feature, weaken the features in the first enhancement feature that are dissimilar to the third enhancement feature, and obtain a high-quality second enhancement feature.

[0140] Based on the technical solutions of the above embodiments, an optional embodiment is provided. In this optional embodiment, the step of determining the third enhancement factor in S520 is refined.

[0141] See also Figure 6 , the steps of determining the third enhancement factor include:

[0142] S610: Determine second similarity data between the features in the first enhanced feature using a preset similarity calculation method.

[0143] The second similarity data may reflect the similarity between the features in the first enhanced feature. The similarity between the features in the first enhanced feature may be accurately determined by using a preset similarity calculation method.

[0144] For example, the preset similarity calculation method uses a Euclidean distance calculation method to calculate the Euclidean distance between each feature in the first enhanced feature to obtain the second similarity data. For example, if the first enhanced feature includes N features, the second similarity data is a Euclidean distance matrix of size N*N.

[0145] It can be seen that the second similarity data reflects the similarity between the features in the first enhanced features.

[0146] S620: Obtain third similarity data corresponding to the third enhanced feature.

[0147] The third similarity data represents the similarity between the features in the third enhanced feature.

[0148] For example, at the previous moment, the third similarity data is the result of calculating the similarity between each feature in the third enhanced feature using the Euclidean distance calculation method. For example, if the third enhanced feature includes N features, the third similarity data is a Euclidean distance matrix of size N*N.

[0149] It can be seen that the third similarity data reflects the similarity between the features in the third enhanced feature.

[0150] The Euclidean distance matrix in the above S610 and S620 adopts a normalized Euclidean distance matrix, that is, the calculated Euclidean distance is normalized by normalization, thereby removing the dimension and facilitating calculation.

[0151] S630: Determine a third enhancement factor according to the second similarity data and the third similarity data.

[0152] In an optional implementation, the process of determining the third enhancement factor in S630 includes: calculating a similarity matrix between the second similarity data and the third similarity data; keeping the values ​​of elements in the similarity matrix that are greater than a preset threshold unchanged, and setting the values ​​of elements in the similarity matrix that are less than or equal to the preset threshold to 0, thereby preprocessing the similarity matrix; and performing dimensionality reduction processing on the preprocessed similarity matrix to obtain the third enhancement factor.

[0153] The preset threshold value can be set as needed, for example, to 0.4. Of course, other values ​​can also be set, which are not limited here.

[0154] It is understandable that preprocessing the similarity matrix can retain the similarity elements corresponding to the time-series related features in the historical vibration data, improve the effectiveness of the third enhancement factor, and in the process of using the third enhancement factor to enhance the first enhancement feature, it helps to capture the hidden time correlation in the vibration data and improve the representation ability of the second enhancement feature.

[0155] It can be seen that the similarity matrix reflects the similarity between the second similarity data and the third similarity data, and indirectly reflects the feature similarity between the first enhanced feature and the third enhanced feature.

[0156] Exemplarily, the second similarity data is an N*N Euclidean distance matrix, the third similarity data is also an N*N Euclidean distance matrix, the similarity matrix between the second similarity data and the third similarity data is N*N in size, and each element in the similarity matrix represents the similarity between the element at a corresponding position in the second similarity data and the element at a corresponding position in the third similarity data. The similarity matrix is ​​reduced in dimension using an activation function (e.g., a Sigmoid function) to obtain third data of size N*1, and the third array is a third enhancement factor.

[0157] It can be seen that based on the second similarity data corresponding to the first enhancement feature and the third similarity data corresponding to the third enhancement feature, the feature similarity between the first enhancement feature and the third enhancement feature can be indirectly known, thereby obtaining the third enhancement factor that characterizes the feature similarity between the first enhancement feature and the third enhancement feature, ensuring the effectiveness of the third enhancement factor, and helping to ensure the effect of using the third enhancement factor to enhance the feature of the first enhancement feature.

[0158] On the basis of the technical solutions of the above embodiments, an optional embodiment is further provided. In this optional embodiment, the feature enhancement step of the first enhanced feature in S530 is refined.

[0159] See also Figure 7 , the feature enhancement step of the first enhancement feature includes:

[0160] S710: Determine a fused vibration feature according to the first enhanced feature and the third enhanced feature.

[0161] Among them, the first enhanced feature reflects the instantaneous feature information of the vibration data at the current moment, and the third enhanced feature reflects the instantaneous feature information of the vibration data at the historical moment. By fusing the first enhanced feature and the third enhanced feature and fusing the vibration features into feature information within a longer time range, the feature evolution process from the historical moment to the current moment can be characterized, and the long-term dependency of the vibration data can be captured.

[0162] The fusion method may be splicing or other fusion methods, which are not limited here.

[0163] S720: Perform feature enhancement on the first enhanced feature according to the fused vibration feature and the third enhancement factor to obtain a second enhanced feature.

[0164] For example, the first enhancement feature and the third enhancement feature are first concatenated to obtain a fused vibration feature. The first enhancement feature is then enhanced using the following formula:

[0165]

[0166] Where, is the second enhanced feature, is the first enhanced feature, is the third enhancement factor, is the activation function (for example, relu activation function), W() is the fully connected layer function, To integrate vibration characteristics.

[0167] From the above calculation formula, we can know that the fused vibration feature is input into the fully connected layer function, and the dimension reduction processing is performed by the fully connected layer function. For example, the size of the first enhanced feature and the third enhanced feature is N*1, and the size of the fused vibration feature is 2N*1. After passing through the fully connected layer function, the size of the fused vibration feature becomes N*1, and then the data is processed by the activation function (for example, if the input data element is greater than 0, the data element is output, and if the data element is less than or equal to 0, 0 is output). The size of the output result of the activation function is still N*1. The size of the third enhancement factor is N*1, and the activation function The size of the output result is N*1, the size of the first enhanced feature is N*1, and the elements at corresponding positions in these three vectors are multiplied to obtain the second enhanced feature with a size of N*1. The features in the second enhanced feature are dimensionless.

[0168] Among them, the fully connected layer function gradually learns which features of the first enhanced feature and the third enhanced feature are more important and thus retained, and which features are unimportant and thus discarded in the process of dimensionality reduction, so that the fused vibration feature after passing through the fully connected layer function retains important features and discards unimportant features, which strengthens the enhancement process of the first enhanced feature, thereby improving the enhancement effect.

[0169] Among them, through the activation function data processing process to achieve further optimization of data.

[0170] It can be understood that the above calculation formula combines the dimensionality reduction processing of the fully connected layer function and the data processing process of the activation function to optimize the fused vibration features, thereby improving the feature enhancement effect.

[0171] It can be seen that when the first enhancement feature is enhanced, in addition to the third enhancement factor, it is also based on the fused vibration feature, and the fused vibration feature is determined based on the first enhancement feature and the third enhancement feature, thereby improving the effect of feature enhancement on the first enhancement feature.

[0172] Based on the technical solutions of the above embodiments, an optional embodiment is also provided. In this optional embodiment, the abnormality determination step of the current vibration data in S150 is refined.

[0173] See also Figure 8 , the abnormality determination steps include:

[0174] S810: Determine the degree of difference between the reconstructed vibration data and the current vibration data.

[0175] The difference degree refers to the overall difference between the reconstructed vibration data and the current vibration data.

[0176] In one optional implementation, the degree of difference is calculated by calculating the sum of the norms of the differences between each vibration signal in the reconstructed vibration data and the corresponding vibration signal in the current vibration data, using this sum as the degree of difference. The norm may be, but is not limited to, a Euclidean norm. This method can determine the overall difference between the reconstructed vibration data and the current vibration data.

[0177] S820: Determine whether the current vibration data is abnormal based on the degree of difference.

[0178] It is understandable that if the degree of difference is greater than a preset threshold, it is determined that the current vibration data has an abnormality; if the degree of difference is less than or equal to the preset threshold, it is determined that the current vibration data has no abnormality.

[0179] As can be seen, through S810 and S820, it is possible to determine whether the current vibration data is abnormal. This allows relevant personnel to be notified when an abnormality occurs, providing an early warning notification. This allows relevant personnel to promptly repair the motor rotor, extending its service life and reducing the probability of failure. This method of determining whether the current vibration data is abnormal based on the degree of difference described above is simple and easy to implement.

[0180] In one embodiment, see Figure 9 , the steps to obtain the initial enhanced features include:

[0181] 1. Input the current vibration data into a non-shared parameter MLP for feature extraction to obtain the current vibration feature vector. Multiply the current vibration feature vector by the transpose of the current vibration feature vector to obtain a first correlation matrix. Calculate the Euclidean distance between each vibration signal in the current vibration data to obtain a second correlation matrix.

[0182] 2. Determine the data fusion coefficient based on the first correlation matrix and the second correlation matrix.

[0183] 3. According to the data fusion coefficient, the first correlation matrix and the second correlation matrix are subjected to feature fusion to obtain degree fusion data.

[0184] 4. Perform dimensionality reduction on the degree fusion data to obtain the first enhancement factor.

[0185] 5. Perform data enhancement on the current vibration data according to the first enhancement factor to obtain enhanced vibration data; input the enhanced vibration data into the first layer of non-shared parameter MLP in the VAE encoder to obtain initial enhanced features.

[0186] In one embodiment, see Figure 10 , the steps of determining the third enhancement factor include:

[0187] 1. Determine the second enhancement factor based on the initial enhancement characteristics.

[0188] 2. The initial enhanced feature is enhanced according to the second enhancement factor to obtain the first enhanced feature, and the second layer shared parameter MLP in the VAE encoder is used to reduce the dimension of the first enhanced feature.

[0189] 3. Obtain third similarity data corresponding to the third enhanced feature; the third enhanced feature is the result of enhancing similar features in the historical vibration data of the historical moment, and the historical moment can be but is not limited to the previous moment.

[0190] 4. Using the Euclidean distance method, determine the second similarity data between each feature in the first enhanced feature after dimensionality reduction.

[0191] 5. Determine a third enhancement factor based on the second similarity data and the third similarity data.

[0192] It is understandable that the parameters involved in the above embodiments are dimensionless.

[0193] It should be understood that, although the steps in the flowcharts of the above embodiments are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the flowcharts of the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily to be performed in sequence, but can be performed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0194] Based on the same inventive concept, embodiments of the present application further provide a motor rotor anomaly detection device for implementing the aforementioned motor rotor anomaly detection method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more motor rotor anomaly detection device embodiments provided below can be found in the aforementioned limitations of the motor rotor anomaly detection method and are not further elaborated here.

[0195] In one embodiment, a device for detecting abnormality of a motor rotor is provided. Figure 11 The motor rotor abnormality detection device includes a data acquisition module 1110, a first enhancement module 1120, a second enhancement module 1130, a data reconstruction module 1140 and an abnormality determination module 1150, wherein:

[0196] The data acquisition module 1110 is used to acquire the current vibration data of the motor rotor at the current moment and the historical vibration data at the historical moments;

[0197] A first enhancement module 1120 is configured to enhance similar features in the current vibration data to obtain first enhanced features;

[0198] The second enhancement module 1130 is configured to enhance the first enhancement feature according to the current vibration data and the historical vibration data to obtain a second enhancement feature;

[0199] A data reconstruction module 1140 is configured to reconstruct data of the second enhanced feature to obtain reconstructed vibration data;

[0200] The abnormality determination module 1150 is configured to determine whether the current vibration data has an abnormality based on the reconstructed vibration data and the current vibration data.

[0201] In one embodiment, the first enhancement module includes:

[0202] A first determining unit is configured to determine a first enhancement factor based on the current vibration data; the first enhancement factor represents the similarity between the vibration signals in the current vibration data;

[0203] a second determining unit, configured to perform feature enhancement on the current vibration data according to the first enhancement factor to obtain an initial enhanced feature;

[0204] A third determining unit is configured to determine a second enhancement factor based on the initial enhancement feature, wherein the second enhancement factor represents the similarity between features in the initial enhancement feature;

[0205] The first enhancement unit is configured to perform feature enhancement on the initial enhancement feature according to the second enhancement factor to obtain a first enhancement feature.

[0206] In one embodiment, the first determining unit includes:

[0207] a first determining subunit, configured to respectively determine first similarity data between different vibration signals in the current vibration data using at least two similarity calculation methods;

[0208] The second determining subunit is configured to determine a first enhancement factor according to each piece of first similarity data.

[0209] In one embodiment, the second determination subunit is specifically used to: determine a data fusion coefficient based on each first similarity data; perform data fusion on each first similarity data based on the data fusion coefficient to obtain degree fusion data; perform dimensionality reduction processing on the degree fusion data to obtain a first enhancement factor.

[0210] In one embodiment, the second enhancement module includes:

[0211] A feature acquisition unit, configured to acquire a third enhanced feature corresponding to the historical vibration data; wherein the third enhanced feature is a result of feature enhancement of similar features in the historical vibration data;

[0212] a fourth determining unit, configured to determine a third enhancement factor according to the first enhancement feature and the third enhancement feature, wherein the third enhancement factor represents feature similarity between the first enhancement feature and the third enhancement feature;

[0213] The second enhancement unit is configured to perform feature enhancement on the first enhancement feature according to a third enhancement factor to obtain a second enhancement feature.

[0214] In an embodiment, the fourth determining unit is specifically configured to: determine second similarity degree data between each feature in the first enhanced feature by using a preset similarity calculation manner; obtain third similarity degree data corresponding to the third enhanced feature; the third similarity degree data represents similarity between each feature in the third enhanced feature; and determine the third enhancement factor according to the second similarity degree data and the third similarity degree data.

[0215] In an embodiment, the second enhancing unit is specifically configured to: determine a fused vibration feature according to the first enhanced feature and the third enhanced feature; and perform feature enhancement on the first enhanced feature to obtain the second enhanced feature according to the fused vibration feature and the third enhancement factor.

[0216] In an embodiment, the anomaly determining module is specifically configured to: determine a difference degree between the reconstructed vibration data and the current vibration data; and determine whether the current vibration data is abnormal according to the difference degree.

[0217] The above-mentioned modules in the motor rotor anomaly detection apparatus can be all or partially implemented by software, hardware, and combinations thereof. The above-mentioned modules can be embedded in or independent of a processor in a computer device in a hardware form, or can be stored in a memory in a computer device in a software form, so as to be called and executed by a processor to perform operations corresponding to the above-mentioned modules.

[0218] In an exemplary embodiment, a computer device, which can be a server, is provided, and an internal structure diagram of the computer device can be as shown in Figure 12 The computer device includes a processor, a memory, an input / output interface, and a communication interface. The processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store related data. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with external terminals through a network connection. The computer program is executed by the processor to implement a motor rotor anomaly detection method.

[0219] Those skilled in the art can understand that, Figure 12The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0220] In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the motor rotor abnormality detection method provided in the above embodiments when executing the computer program.

[0221] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the motor rotor abnormality detection method provided in the above embodiments is implemented.

[0222] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the motor rotor abnormality detection method provided in the above embodiments is implemented.

[0223] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0224] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), quantum computing-based data processing logic devices, artificial intelligence (AI) processors, and the like.

[0225] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0226] The above embodiments merely illustrate several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for detecting abnormality of a motor rotor, characterized in that: include: Obtain the current vibration data of the motor rotor at the current moment and the historical vibration data at the historical moments; Performing feature enhancement on similar features in the current vibration data to obtain a first enhanced feature; Performing feature enhancement on the first enhanced feature according to the current vibration data and the historical vibration data to obtain a second enhanced feature; reconstructing data of the second enhanced feature to obtain reconstructed vibration data; It is determined whether the current vibration data has an abnormality according to the reconstructed vibration data and the current vibration data.

2. The method according to claim 1, characterized in that The step of enhancing the similar features in the current vibration data to obtain a first enhanced feature includes: determining a first enhancement factor based on the current vibration data, wherein the first enhancement factor represents the similarity between vibration signals in the current vibration data; Performing feature enhancement on the current vibration data according to the first enhancement factor to obtain an initial enhanced feature; Determining a second enhancement factor based on the initial enhancement feature, wherein the second enhancement factor represents the similarity between each feature in the initial enhancement feature; The initial enhanced feature is enhanced according to the second enhancement factor to obtain the first enhanced feature.

3. The method according to claim 2, characterized in that The determining of a first enhancement factor according to the current vibration data includes: using at least two similarity calculation methods to respectively determine first similarity data between different vibration signals in the current vibration data; The first enhancement factor is determined according to each of the first similarity data.

4. The method according to claim 3, characterized in that The determining the first enhancement factor according to each of the first similarity data includes: determining a data fusion coefficient according to each of the first similarity data; performing data fusion on each of the first similarity degree data according to the data fusion coefficient to obtain degree fusion data; Performing dimensionality reduction processing on the degree fusion data to obtain the first enhancement factor.

5. The method according to claim 1, wherein The step of enhancing the first enhanced feature according to the current vibration data and the historical vibration data to obtain a second enhanced feature includes: Acquire a third enhanced feature corresponding to the historical vibration data; wherein the third enhanced feature is a result of enhancing similar features in the historical vibration data; Determining a third enhancement factor according to the first enhancement feature and the third enhancement feature; wherein the third enhancement factor represents feature similarity between the first enhancement feature and the third enhancement feature; The first enhancement feature is enhanced according to the third enhancement factor to obtain the second enhancement feature.

6. The method according to claim 5, characterized in that The determining a third enhancement factor according to the first enhancement feature and the third enhancement feature includes: Determining second similarity data between each feature in the first enhanced feature using a preset similarity calculation method; Obtaining third similarity data corresponding to the third enhanced feature; the third similarity data represents the similarity between each feature in the third enhanced feature; The third enhancement factor is determined according to the second similarity data and the third similarity data.

7. The method according to claim 5, characterized in that The step of performing feature enhancement on the first enhancement feature according to the third enhancement factor to obtain the second enhancement feature includes: determining a fused vibration feature according to the first enhanced feature and the third enhanced feature; The first enhanced feature is enhanced according to the fused vibration feature and the third enhancement factor to obtain the second enhanced feature.

8. The method according to any one of claims 1 to 7, characterized in that The determining, based on the reconstructed vibration data and the current vibration data, whether the current vibration data is abnormal includes: determining a degree of difference between the reconstructed vibration data and the current vibration data; According to the degree of difference, it is determined whether the current vibration data is abnormal.

9. A motor rotor abnormality detection device, characterized in that: include: A data acquisition module is used to acquire the current vibration data of the motor rotor at the current moment and the historical vibration data at the historical moments; A first enhancement module is configured to enhance similar features in the current vibration data to obtain first enhanced features; a second enhancement module, configured to perform feature enhancement on the first enhancement feature according to the current vibration data and the historical vibration data to obtain a second enhancement feature; a data reconstruction module, configured to reconstruct data of the second enhanced feature to obtain reconstructed vibration data; The abnormality determination module is configured to determine whether the current vibration data has an abnormality based on the reconstructed vibration data and the current vibration data.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images