Motor magnetic steel sheet defect detection method based on vibration analysis

Abstract

The invention provides a motor magnetic steel sheet defect detection method based on vibration analysis, and relates to the technical field of motor fault diagnosis, and the method comprises the steps: carrying out the vibration feature extraction of a plurality of vibration data sets, inputting a plurality of vibration feature sets into a defect feature recognition library, carrying out the similar traversal comparison, and outputting a plurality of initial defect features; and collecting temperature distribution characteristics and sound wave characteristics of the target motor in a predetermined period, carrying out credible probability analysis on the plurality of initial defect characteristics, and outputting the initial defect characteristic with the maximum probability as a motor magnetic steel sheet defect detection result. According to the method and the device, the technical problems that a traditional motor magnetic steel sheet defect detection method cannot quickly and accurately identify the defect characteristics of the motor magnetic steel sheet during the operation of the motor, and the defect detection timeliness and accuracy are insufficient can be solved; and the efficiency, the accuracy and the reliability of defect detection of the magnetic steel sheet of the motor can be remarkably improved.

Description

Technical Field

[0001] The present application relates to the technical field of motor fault diagnosis, and in particular to a method for detecting defects in motor magnetic steel sheets based on vibration analysis. Background Art

[0002] As an important power equipment in industrial production, the health status of the motor directly affects the equipment's operating efficiency, production safety and maintenance costs.

[0003] As an important part of the motor, the motor magnetic steel sheet is subjected to high-frequency magnetic fields and mechanical stress for a long time, and is prone to defects such as cracks, shedding, and wear. These defects can cause unbalanced operation of the motor, increased vibration, reduced efficiency, and even cause major failures, resulting in downtime and equipment damage. Therefore, timely and accurate detection of defects in the motor magnetic steel sheet is crucial to ensure the safe operation of the motor. Summary of the invention

[0004] The purpose of this application is to provide a motor magnetic steel sheet defect detection method based on vibration analysis, so as to solve the technical problems that the traditional motor magnetic steel sheet defect detection method cannot quickly and accurately identify the defect characteristics of the motor magnetic steel sheet when the motor is running, and there is a lack of timeliness and accuracy in defect detection.

[0005] In view of the above problems, the present application provides a motor magnetic steel sheet defect detection method based on vibration analysis, including: building a vibration monitoring array based on predetermined monitoring points, continuously monitoring the target motor through the vibration monitoring array, and obtaining multiple vibration data sets within a predetermined period; performing vibration feature extraction on the multiple vibration data sets to obtain multiple vibration feature sets; respectively inputting the multiple vibration feature sets into a defect feature recognition library for similarity traversal comparison, selecting multiple defect feature sets that meet the dynamic comparison threshold, and performing intersection analysis on the multiple defect feature sets to output multiple initial defect features; collecting the temperature distribution characteristics and acoustic wave characteristics of the target motor within a predetermined period, performing credible probability analysis on the multiple initial defect features according to the temperature distribution characteristics and acoustic wave characteristics, and outputting the initial defect feature with the maximum probability as the motor magnetic steel sheet defect detection result of the target motor.

[0006] The technical solution provided in this application has at least the following technical effects or advantages: A vibration monitoring array is built based on predetermined monitoring points, and the target motor is continuously monitored through the vibration monitoring array to obtain multiple vibration data sets within a predetermined period; then, vibration features are extracted from the multiple vibration data sets to obtain multiple vibration feature sets; further, the multiple vibration feature sets are respectively input into the defect feature recognition library for similarity traversal comparison, and multiple defect feature sets that meet the dynamic comparison threshold are selected, and intersection analysis is performed on the multiple defect feature sets to output multiple initial defect features; on the other hand, the temperature distribution features and acoustic wave features of the target motor within a predetermined period are collected; then, the multiple initial defect features are analyzed for credible probability based on the temperature distribution features and acoustic wave features, and the initial defect feature with the maximum probability is output as the motor magnetic steel sheet defect detection result of the target motor; that is, through multi-point vibration monitoring similarity comparison, multiple related potential defect features can be quickly discovered, and then the multiple related potential defect features are jointly analyzed and credibility judged through temperature and acoustic wave features to determine the defect feature with the highest credibility, thereby significantly improving the efficiency, accuracy and reliability of motor magnetic steel sheet defect detection.

[0007] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0009] Figure 1 A schematic flow chart of a method for detecting defects in motor magnetic steel sheets based on vibration analysis according to the present application.

[0010] Figure 2 A schematic diagram of a flow chart of obtaining multiple vibration feature sets in a method for detecting defects in motor magnetic steel sheets based on vibration analysis in the present application. DETAILED DESCRIPTION

[0011] This application provides a motor magnetic steel sheet defect detection method based on vibration analysis, which solves the technical problem that the traditional motor magnetic steel sheet defect detection method cannot quickly and accurately identify the defect characteristics of the motor magnetic steel sheet when the motor is running, and there are insufficient timeliness and accuracy of defect detection. Through multi-point vibration monitoring similarity comparison, multiple related potential defect characteristics can be quickly discovered, and then the temperature and sound wave characteristics are used to jointly analyze and determine the credibility of multiple related potential defect characteristics, and determine the defect characteristics with the highest credibility, thereby significantly improving the efficiency, accuracy and reliability of motor magnetic steel sheet defect detection.

[0012] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0013] For examples, please see the attached Figure 1 The present application provides a method for detecting defects in motor magnetic steel sheets based on vibration analysis, which specifically includes the following steps: S100: Building a vibration monitoring array based on predetermined monitoring points, continuously monitoring a target motor through the vibration monitoring array, and acquiring a plurality of vibration data sets within a predetermined period.

[0014] Furthermore, step S100 of the present application also includes: S110: Obtain predetermined monitoring points of the target motor, wherein the predetermined monitoring points include at least a motor stator, a motor rotor and a motor bearing; S120: Arrange vibration sensors at the motor stator, the motor rotor and the motor bearing respectively to construct a vibration monitoring array; S130: Continuously monitor the target motor through the vibration monitoring array at a predetermined monitoring frequency to obtain multiple vibration data sets of multiple monitoring points within a predetermined period.

[0015] Specifically, first, the predetermined monitoring points of the target motor are obtained, wherein the predetermined monitoring points include at least the motor stator, the motor rotor and the motor bearing. These monitoring points cover the core components of the motor and can effectively reflect the overall operating status of the motor and the related signals of the magnetic steel sheet defects. Then, vibration sensors are arranged at the motor stator, the motor rotor and the motor bearing respectively, and each vibration sensor is connected to the data acquisition system to construct a comprehensive vibration monitoring array, which ensures that the vibration signals of each key part can be collected and analyzed in real time by coordinating the work of each sensor. Then, according to the operating characteristics of the motor, a suitable monitoring frequency is set (such as collecting a certain number of vibration data per second) to ensure that all vibration changes in the motor operation process can be captured, especially weak vibration signals related to magnetic steel sheet defects; finally, according to the predetermined monitoring frequency, the target motor is continuously monitored through the vibration monitoring array to obtain multiple vibration data sets of multiple monitoring points (motor stator, motor rotor and motor bearing) within a predetermined period (such as within the next 10 minutes). By obtaining multiple vibration data sets of multiple monitoring points, basic data support is provided for subsequent defect analysis.

[0016] S200: Extract vibration features from the multiple vibration data sets to obtain multiple vibration feature sets.

[0017] Further, if Figure 2 As shown, step S200 of the present application also includes: S210: performing noise point analysis and filtering and denoising processing on the multiple vibration data sets to obtain multiple noise point ratios and multiple standard vibration data sets; S220: configuring vibration characteristic indicators, wherein the vibration characteristic indicators at least include abnormal frequency components, abnormal frequency proportions, amplitude curves and waveform curves; S230: performing vibration feature extraction on the multiple standard vibration data sets according to the vibration characteristic indicators to obtain multiple vibration feature sets for multiple monitoring points.

[0018] Specifically, first, the multiple vibration data sets are subjected to noise point analysis and filtering and denoising processing respectively, such as using statistical analysis methods (such as mean and standard deviation analysis) to identify abnormal fluctuations or noise points in the data. Common noise points include sensor failures, abnormal signals caused by external interference, etc. The higher the noise point ratio, the more noise components there are in the signal; a suitable filtering algorithm is used to filter the vibration data to remove high-frequency noise or low-frequency interference. Common filtering methods include wavelet transform, Kalman filtering, mean filtering, etc.; multiple noise point ratios and multiple standard vibration data sets are obtained, wherein the noise point ratio reflects the relative influence of noise in the data, and the standard vibration data set is a clear signal after denoising.

[0019] Next, vibration characteristic indicators are configured, wherein the vibration characteristic indicators are used for subsequent feature extraction and fault analysis. These characteristic indicators can effectively capture abnormal changes and potential fault signs in vibration signals, and at least include abnormal frequency components, abnormal frequency proportion, amplitude curve and waveform curve. Abnormal frequency components are usually related to motor faults (such as cracks and detachment of magnetic steel sheets). The frequency components of the vibration signal are extracted by Fourier transform or fast Fourier transform (FFT), and specific abnormal frequency components (such as 1X, 2X, 3X and other frequency multiplication characteristics) are identified and compared with the frequency mode of the normal motor; the abnormal frequency proportion is the proportion of the abnormal frequency component in the entire frequency spectrum, which is used to indicate the severity of the abnormality. The higher the proportion, the greater the possibility of fault; the amplitude curve can intuitively show the change of vibration intensity of the motor in different time periods, and abnormal amplitude changes are often related to internal faults or imbalances of the motor; the waveform curve reflects the change of the vibration signal over time. By analyzing the shape of the waveform curve, the fault mode inside the motor can be identified.

[0020] Then, according to the vibration characteristic indicators, vibration characteristics are extracted from the multiple standard vibration data sets, such as using Fourier transform (FFT) and other methods to extract the frequency components of the vibration signal and identify abnormal frequency components and corresponding amplitudes; analyzing the changes in the vibration signal through time domain statistical characteristics (such as peak value, root mean square value, standard deviation, etc.) to identify possible vibration anomalies; performing morphological analysis on the waveform of the vibration signal to detect whether there are irregular vibration waveforms; then combining time domain, frequency domain and waveform analysis to extract multi-dimensional vibration characteristics of each monitoring point (stator, rotor, bearing, etc.) to obtain multiple vibration feature sets for multiple monitoring points, wherein the vibration feature sets include indicators such as abnormal frequency components, abnormal frequency proportions, amplitude curves and waveform curves. These feature sets provide a detailed basis for subsequent defect identification and diagnosis.

[0021] S300: inputting the multiple vibration feature sets into the defect feature recognition library for similarity traversal comparison respectively, selecting multiple defect feature sets that meet the dynamic comparison threshold, performing intersection analysis on the multiple defect feature sets, and outputting multiple initial defect features.

[0022] Furthermore, step S300 of the present application also includes: S310: Constructing a defect feature recognition library based on the motor property features, the motor magnetic steel sheet property features and the vibration feature index.

[0023] Furthermore, step S310 of the present application also includes: S311: Taking the motor attribute characteristics and the motor magnetic steel sheet attribute characteristics as equipment constraints, taking the vibration characteristic index as conditional constraints, taking the motor magnetic steel sheet defect detection based on vibration analysis as a guide, performing information retrieval based on big data, obtaining multiple sample vibration feature sets, and marking the defect types and defect intensities of the motor magnetic steel sheets under different sample vibration feature sets, obtaining multiple sample defect types and multiple sample defect intensities; S312: Clustering the multiple sample defect types and multiple sample defect intensities based on the same defect type and defect intensity interval, determining multiple standard defect types, wherein each standard defect type is marked with multiple defect intensity intervals; S313: Randomly selecting a first standard defect type, and a first defect intensity interval of the first standard defect type; S314: performing high-frequency feature analysis on the multiple first sample vibration feature sets to determine a first standard vibration feature set, wherein the first standard vibration feature set includes a first standard abnormal frequency component, a first standard abnormal frequency ratio, a first standard amplitude curve and a first standard waveform curve; S315: establishing a first mapping association between the first standard vibration feature set and the first standard defect type and the first defect intensity interval, and constructing a first defect identification branch according to the first mapping association; S316: sequentially analyzing and obtaining multiple defect identification branches, and constructing the defect feature identification library according to the multiple defect identification branches.

[0024] Specifically, first, the motor attribute characteristics and the motor magnetic steel sheet attribute characteristics are obtained, wherein the motor attribute characteristics include the motor type, operating status, motor component characteristics, etc.; the motor magnetic steel sheet attribute characteristics include material characteristics, magnetic steel sheet shape and size, etc. Then, the motor attribute characteristics and the motor magnetic steel sheet attribute characteristics are used as equipment constraints, that is, the collection range and type of vibration characteristics are limited by equipment constraints, the vibration characteristic index is used as a conditional constraint, and the motor magnetic steel sheet defect detection based on vibration analysis is used as a guide. Information retrieval is performed based on big data, that is, based on the motor operating status, the monitoring data of the vibration sensor, and the different working conditions of the motor magnetic steel sheet, the vibration data sets of multiple motors are collected through the data acquisition system; and after data cleaning and preprocessing using the big data analysis platform, the vibration characteristics that meet the equipment constraints and conditional constraints are extracted to form multiple sample vibration feature sets. The defect types and defect intensities of the motor magnetic steel sheets under different sample vibration feature sets are further labeled, that is, each vibration feature set is manually labeled or classified through an intelligent analysis algorithm (such as a machine learning algorithm), and the defect types (such as cracks, shedding, warping, wear, etc.) of the motor magnetic steel sheets and their corresponding defect intensities (mild, moderate, severe) are labeled; multiple sample defect types and multiple sample defect intensities are obtained.

[0025] Next, the multiple sample defect types and multiple sample defect intensities are clustered based on the same defect type and defect intensity interval. For example, a clustering algorithm (such as K-means, DBSCAN, hierarchical clustering, etc.) is used to perform cluster analysis on the defect type and intensity. The purpose of this step is to classify samples with similar characteristics into the same category, thereby identifying multiple standard defect types and their corresponding defect intensity intervals. The clustering analysis maps the relationship between defect type and defect intensity to ensure that defects of the same type are clustered together, and different intensity intervals are divided according to the change in defect intensity to determine multiple standard defect types (such as cracks, detachment, warping, and wear), where each standard defect type is marked with multiple defect intensity intervals.

[0026] Then, any one defect type is randomly selected from the multiple standard defect types as the first standard defect type, and the first defect intensity interval (any intensity interval) of the first standard defect type is randomly selected; then, the multiple sample vibration feature sets are screened according to the first standard defect type and the first defect intensity interval to determine multiple first sample vibration feature sets. The multiple first sample vibration feature sets are further subjected to high-frequency feature analysis, and high-frequency components related to defects are extracted by performing high-frequency feature analysis on the multiple first sample vibration feature sets. These frequency components can usually reflect the nature and intensity of the defects, such as selecting the mode value of the multiple first sample vibration feature sets as the first standard vibration feature set, and obtaining the first standard vibration feature set, wherein the first standard vibration feature set includes the first standard abnormal frequency component, the first standard abnormal frequency ratio, the first standard amplitude curve, and the first standard waveform curve.

[0027] Further establish a first mapping association between the first standard vibration feature set and the first standard defect type and the first defect intensity interval. This mapping relationship associates different vibration features with the specific type and intensity of the defect, providing a basis for subsequent defect diagnosis. Then, based on the decision tree principle, with the first standard vibration feature set as a child node, and the first standard defect type and the first defect intensity interval as leaf nodes of the child node, a first defect identification branch is constructed according to the first mapping association. This branch is based on the conditional constraints of the first standard vibration feature set and the defect type intensity, and can accurately identify defects for specific defect types and intensities. Use the same method to analyze and construct multiple defect identification branches in sequence, and then integrate the multiple defect identification branches into a complete defect feature identification library, which contains different defect types, different defect intensities and their corresponding vibration features, thereby providing a comprehensive basis for online defect detection of motor magnetic steel sheets.

[0028] S320: Calculate the mean of the multiple noise point ratios to determine the comprehensive noise coefficient, calculate the ratio of the comprehensive noise coefficient to the standard noise coefficient, and obtain a threshold compensation coefficient; S330: Compensate the initial similarity comparison threshold according to the threshold compensation coefficient to obtain a dynamic comparison threshold.

[0029] Specifically, in vibration data, noise points may affect the accuracy of vibration feature extraction and defect feature identification, so it is necessary to analyze the noise points and calculate the comprehensive noise coefficient to evaluate the impact of noise and compensate for it in subsequent processing. First, the average of the multiple noise point ratios is calculated, and the average calculation result is set as the comprehensive noise coefficient. Then, the ratio of the comprehensive noise coefficient to the standard noise coefficient (the ideal noise level obtained based on historical data or theoretical analysis) is calculated to obtain the threshold compensation coefficient. The compensation coefficient reflects the difference between the actual noise and the standard noise, and can be used to adjust the comparison threshold. The initial similarity comparison threshold is further compensated according to the threshold compensation coefficient, that is, the reciprocal of the threshold compensation coefficient is multiplied by the initial similarity comparison threshold. For example, assuming that the threshold compensation coefficient is 0.95 and the initial similarity comparison threshold is 80%, indicating that the data quality is relatively high, and therefore the similarity comparison threshold needs to be increased, then the reciprocal of 0.95 multiplied by 80% is approximately equal to 0.84%, which is set as the dynamic comparison threshold; the dynamic comparison threshold is obtained, wherein the larger the reciprocal of the threshold compensation coefficient is, the slightly higher the dynamic comparison threshold will be, making defect identification more stringent and reducing the possibility of false alarms; the smaller the reciprocal of the threshold compensation coefficient is, the appropriately lower the dynamic comparison threshold will be, making defect identification more tolerant and reducing missed alarms. This dynamic comparison threshold adjustment method can effectively cope with noise changes during motor operation and improve the flexibility and accuracy of defect detection. When the noise environment is more complex, automatic adjustment of the comparison threshold ensures that the system can both guarantee accuracy and avoid missed detections due to overly strict thresholds. When the data quality is high, appropriately increasing the comparison threshold can more accurately identify potential defects and ensure the accuracy and reliability of the detection results.

[0030] S340: Input the multiple vibration feature sets into the defect feature recognition library respectively for similarity traversal comparison, select defect features whose similarity meets the dynamic comparison threshold, and output the multiple defect feature sets, wherein the defect features include the defect type and defect intensity of the motor magnetic steel sheet, and the defect type includes at least cracks, detachment, warping, and wear.

[0031] Furthermore, step S340 of the present application also includes: S341: randomly selecting a first vibration feature set, and randomly selecting a first defect recognition branch in the defect feature recognition library.

[0032] Specifically, first, any one vibration feature set is randomly selected from the multiple vibration feature sets as a first vibration feature set, and a first defect recognition branch is randomly selected from the multiple defect recognition branches of the defect feature recognition library.

[0033] S342: Input the first vibration feature set into the first defect recognition branch, and perform a similarity traversal comparison with the first standard vibration feature set. If the comparison similarities satisfy the dynamic comparison threshold, output the first standard defect type and the first defect intensity range as the first defect feature.

[0034] Furthermore, step S342 of the present application also includes: S3421: Acquire the first vibration feature set, wherein the first vibration feature set includes a first abnormal frequency component, a first abnormal frequency ratio, a first amplitude curve and a first waveform curve; S3422: Perform a similarity comparison between the first abnormal frequency component and the first standard abnormal frequency component to determine a first similarity; S3423: If the first similarity is greater than the dynamic comparison threshold, perform a similarity comparison between the first abnormal frequency ratio and the first standard abnormal frequency ratio to determine a second similarity; S3424: If the second similarity is greater than the dynamic comparison threshold, perform a similarity comparison between the first amplitude curve and the first standard amplitude curve to determine a third similarity; S3425: If the third similarity is greater than the dynamic comparison threshold, perform a similarity comparison between the first waveform curve and the first standard waveform curve to determine a fourth similarity, and if the fourth similarity is greater than the dynamic comparison threshold, set the output first standard defect type and the first defect intensity interval as the first defect feature.

[0035] Specifically, the first vibration feature set is obtained, wherein the first vibration feature set includes a first abnormal frequency component, a first abnormal frequency ratio, a first amplitude curve and a first waveform curve; then, a similarity comparison is performed on the first abnormal frequency component and the first standard abnormal frequency component, such as by calculating the Euclidean distance for similarity comparison, to obtain a first similarity; if the first similarity is greater than the dynamic comparison threshold, a similarity comparison is performed on the first abnormal frequency ratio and the first standard abnormal frequency ratio to determine a second similarity; if the second similarity is greater than the dynamic comparison threshold, a similarity comparison is performed on the first amplitude curve and the first standard amplitude curve to determine a third similarity; if the third similarity is greater than the dynamic comparison threshold, a similarity comparison is performed on the first waveform curve and the first standard waveform curve to determine a fourth similarity, and if the fourth similarity is greater than the dynamic comparison threshold, the output first standard defect type (such as cracks, shedding, wear, etc.) and the first defect intensity interval are set as the first defect feature. If the first similarity is less than or equal to the dynamic comparison threshold, the second similarity is less than or equal to the dynamic comparison threshold, the third similarity is less than or equal to the dynamic comparison threshold, or the fourth similarity is less than or equal to the dynamic comparison threshold, the comparison is stopped.

[0036] S343: Perform similarity traversal comparison on the first vibration feature set and standard vibration feature sets of other defect recognition branches in the defect feature recognition library in turn, obtain a first defect feature set, and add it to the multiple defect feature sets.

[0037] Specifically, using the same method, the first vibration feature set is sequentially traversed and compared with the standard vibration feature sets of other defect recognition branches in the defect feature recognition library, and the first defect feature set that meets the dynamic comparison threshold is output, and multiple defect feature sets of multiple monitoring points are obtained in turn. Finally, the multiple defect feature sets are subjected to intersection analysis, and multiple initial defect features are output. Through intersection analysis, defect features that are commonly supported by multiple monitoring points and different feature sets can be found, noise and redundant features can be removed, and the most representative defect features can be extracted, which significantly improves the accuracy, reliability and robustness of motor magnetic steel sheet defect detection.

[0038] S400: Collect the temperature distribution characteristics and acoustic wave characteristics of the target motor within a predetermined period, perform a credible probability analysis on the multiple initial defect characteristics according to the temperature distribution characteristics and acoustic wave characteristics, and output the initial defect characteristic with the maximum probability as the motor magnetic steel sheet defect detection result of the target motor.

[0039] Furthermore, step S400 of the present application also includes: S410: Taking the motor attribute characteristics and the motor magnetic steel sheet attribute characteristics as equipment constraints, taking the defect type and defect intensity as feature constraints, and taking the motor magnetic steel sheet defect detection as a guide, retrieve and obtain a sample temperature distribution feature set, a sample acoustic wave feature set, and multiple sample defect feature sets, and count the proportions of sample defect features under different sample temperature distribution characteristics and different sample acoustic wave characteristics, set them as the triggering probability of multiple defect features, construct a sample defect feature probability distribution, and obtain a sample defect feature probability distribution set; S420: Using the sample temperature distribution feature set, the sample acoustic wave feature set, the multiple sample defect feature sets, and the sample defect feature probability distribution set to N Prediction operators are used for supervised training until convergence to obtain N defect feature trigger probability prediction units, and a defect feature trigger probability predictor is integrated and constructed based on the N defect feature trigger probability prediction units, wherein the prediction operator includes at least a BP neural network, a random forest and a support vector machine, and N is an integer greater than or equal to 3; S430: the temperature distribution characteristics, the acoustic wave characteristics and a plurality of initial defect characteristics are input into the defect feature trigger probability predictor for credible probability analysis, the defect feature probability distribution is output, and the initial defect feature with the maximum probability in the defect feature probability distribution is selected as the motor magnetic steel sheet defect detection result of the target motor.

[0040] Specifically, in the defect detection of motor magnetic steel sheets, in addition to vibration data, temperature distribution characteristics and acoustic wave characteristics are also very important auxiliary information. By combining these characteristics, the accuracy and reliability of defect detection can be further improved. First, the temperature distribution characteristics and acoustic wave characteristics of the target motor within a predetermined period are collected. For example, within a predetermined period (for example, every hour or every minute), the temperature data of the target motor at each monitoring point is collected to generate a temperature distribution feature set. The temperature change may be related to the defects of the motor magnetic steel sheet (such as thermal expansion, friction, etc.), so the temperature characteristics are helpful to detect the existence of defects; an acoustic wave sensor is arranged outside the motor or at other suitable locations to collect the acoustic wave signals generated during the operation of the motor. Defects in the motor magnetic steel sheet are often accompanied by abnormal sound changes, such as collision, friction or vibration, which will be reflected in the acoustic wave signal; the acoustic wave data of the target motor during operation is obtained by the acoustic wave sensor to generate an acoustic wave feature set. The acoustic wave features can be extracted by analyzing parameters such as the acoustic wave spectrum, intensity, and waveform.

[0041] Next, taking the motor property characteristics and the motor magnetic steel sheet property characteristics as equipment constraints, the defect type and defect intensity as feature constraints, and the motor magnetic steel sheet defect detection as a guide, the sample temperature distribution feature set, the sample acoustic wave feature set and multiple sample defect feature sets are obtained based on big data retrieval; further statistics are made on the proportion of sample defect features under different sample temperature distribution characteristics and different sample acoustic wave characteristics, and the feature proportion is set as the trigger probability of the defect feature to obtain the trigger probability of multiple defect features, and the sample defect feature probability distribution is constructed by combining multiple sample defect features to obtain the sample defect feature probability distribution set.

[0042] N prediction operators are configured, wherein the prediction operators include at least BP neural network, random forest and support vector machine, N is an integer greater than or equal to 3, and the value of N can be set according to the actual situation; then the sample temperature distribution feature set, sample acoustic wave feature set, multiple sample defect feature sets and sample defect feature probability distribution set are used as sample training data, and the sample training data is divided into N parts, and the N parts of data are selected N times with replacement to construct the first training set, and iteratively selected N times to obtain N training sets. Then, the sample temperature distribution features, sample acoustic wave features and multiple sample defect features are used as inputs, and the sample defect feature probability distribution is used as output. The BP neural network, random forest and support vector machine are supervised and trained respectively by the N training sets until convergence, and N defect feature trigger probability prediction units are obtained; then, the N defect feature trigger probability prediction units are verified by the verification set, and the N prediction accuracies of the N defect feature trigger probability prediction units are obtained. Further, N output weights are configured according to the N prediction accuracies, wherein the weight is positively correlated with the prediction accuracy, that is, the greater the prediction accuracy, the greater the weight. Finally, the N defect feature trigger probability prediction units are integrated and fused according to the N output weights to construct a defect feature trigger probability predictor, where the output of the defect feature trigger probability predictor is the weighted calculation result of the N defect feature trigger probability prediction units. Through this integrated learning method, the advantages of multiple prediction operators can be fully utilized, and the accuracy of prediction can be improved by weighted fusion. The output results of different prediction operators are given different weights according to their performance on the validation set. Therefore, in practical applications, the final defect prediction results are more accurate and robust, and can cope with the task of motor magnetic steel sheet defect detection under complex working conditions.

[0043] Then, the temperature distribution feature, the acoustic wave feature and the multiple initial defect features are input into the defect feature trigger probability predictor for credible probability analysis, and the defect feature probability distribution is output; finally, the initial defect feature with the maximum probability in the defect feature probability distribution is selected as the motor magnetic steel sheet defect detection result of the target motor. Through this scheme, the defect type and intensity of the motor magnetic steel sheet can be identified in real time and efficiently, significantly improving the accuracy and reliability of the motor magnetic steel sheet defect detection.

[0044] In summary, the motor magnetic steel sheet defect detection method based on vibration analysis provided by the present application has the following technical effects: A vibration monitoring array is built based on predetermined monitoring points, and the target motor is continuously monitored through the vibration monitoring array to obtain multiple vibration data sets within a predetermined period; then, vibration features are extracted from the multiple vibration data sets to obtain multiple vibration feature sets; further, the multiple vibration feature sets are respectively input into the defect feature recognition library for similarity traversal comparison, and multiple defect feature sets that meet the dynamic comparison threshold are selected, and intersection analysis is performed on the multiple defect feature sets to output multiple initial defect features; on the other hand, the temperature distribution features and acoustic wave features of the target motor within a predetermined period are collected; then, the multiple initial defect features are analyzed for credible probability based on the temperature distribution features and acoustic wave features, and the initial defect feature with the maximum probability is output as the motor magnetic steel sheet defect detection result of the target motor; that is, through multi-point vibration monitoring similarity comparison, multiple related potential defect features can be quickly discovered, and then the multiple related potential defect features are jointly analyzed and credibility judged through temperature and acoustic wave features to determine the defect feature with the highest credibility, thereby significantly improving the efficiency, accuracy and reliability of motor magnetic steel sheet defect detection.

[0045] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0046] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A method for detecting defects in motor magnetic steel sheets based on vibration analysis, characterized in that: Methods include: Building a vibration monitoring array based on predetermined monitoring points, continuously monitoring the target motor through the vibration monitoring array, and acquiring multiple vibration data sets within a predetermined period; Extracting vibration features from the multiple vibration data sets to obtain multiple vibration feature sets; Input the multiple vibration feature sets into the defect feature recognition library for similarity traversal comparison, select multiple defect feature sets that meet the dynamic comparison threshold, perform intersection analysis on the multiple defect feature sets, and output multiple initial defect features; The temperature distribution characteristics and acoustic wave characteristics of the target motor within a predetermined period are collected, and a credible probability analysis is performed on the multiple initial defect characteristics according to the temperature distribution characteristics and the acoustic wave characteristics, and the initial defect characteristic with the maximum probability is output as the motor magnetic steel sheet defect detection result of the target motor.

2. The method for detecting defects in motor magnetic steel sheets based on vibration analysis according to claim 1, characterized in that: A vibration monitoring array is built based on predetermined monitoring points, and the target motor is continuously monitored through the vibration monitoring array to obtain multiple vibration data sets within a predetermined period, including: Acquire predetermined monitoring points of the target motor, wherein the predetermined monitoring points at least include a motor stator, a motor rotor, and a motor bearing; Vibration sensors are respectively arranged at the motor stator, motor rotor and motor bearing to construct a vibration monitoring array; According to a predetermined monitoring frequency, the target motor is continuously monitored by the vibration monitoring array to obtain a plurality of vibration data sets at a plurality of monitoring points within a predetermined period.

3. The method for detecting defects in motor magnetic steel sheets based on vibration analysis according to claim 2, characterized in that: Vibration feature extraction is performed on the multiple vibration data sets to obtain multiple vibration feature sets, including: Performing noise point analysis and filtering and denoising processing on the multiple vibration data sets to obtain multiple noise point ratios and multiple standard vibration data sets; Configuring vibration characteristic indicators, wherein the vibration characteristic indicators at least include abnormal frequency components, abnormal frequency proportions, amplitude curves, and waveform curves; According to the vibration feature index, vibration feature extraction is performed on the multiple standard vibration data sets to obtain multiple vibration feature sets of multiple monitoring points.

4. The method for detecting defects in motor magnetic steel sheets based on vibration analysis according to claim 3 is characterized in that: The multiple vibration feature sets are respectively input into the defect feature recognition library for similarity traversal comparison, and multiple defect feature sets that meet the dynamic comparison threshold are selected, including: Building a defect feature recognition library based on the motor property features, the motor magnetic steel sheet property features and the vibration feature index; Performing mean calculation on the multiple noise point ratios to determine a comprehensive noise coefficient, calculating a ratio of the comprehensive noise coefficient to a standard noise coefficient, and obtaining a threshold compensation coefficient; Compensating the initial similarity comparison threshold according to the threshold compensation coefficient to obtain a dynamic comparison threshold; The multiple vibration feature sets are respectively input into the defect feature recognition library for similarity traversal comparison, and the defect features whose similarity meets the dynamic comparison threshold are selected, and the multiple defect feature sets are output, wherein the defect features include the defect type and defect intensity of the motor magnetic steel sheet, and the defect type includes at least cracks, shedding, warping, and wear.

5. The method for detecting defects of motor magnetic steel sheets based on vibration analysis according to claim 4, characterized in that: A defect feature recognition library is constructed based on the motor property features, the motor magnetic steel sheet property features and the vibration feature index, including: Taking the motor property characteristics and the motor magnetic steel sheet property characteristics as equipment constraints, taking the vibration characteristic index as conditional constraints, taking the motor magnetic steel sheet defect detection based on vibration analysis as a guide, performing information retrieval based on big data, obtaining multiple sample vibration feature sets, and marking the defect types and defect intensities of the motor magnetic steel sheets under different sample vibration feature sets, obtaining multiple sample defect types and multiple sample defect intensities; Clustering the multiple sample defect types and the multiple sample defect intensities based on the same defect type and defect intensity interval to determine multiple standard defect types, wherein each standard defect type is identified with multiple defect intensity intervals; Randomly selecting a first standard defect type and a first defect intensity interval of the first standard defect type, screening the plurality of sample vibration feature sets according to the first standard defect type and the first defect intensity interval, and determining a plurality of first sample vibration feature sets; Performing high-frequency feature analysis on the plurality of first sample vibration feature sets to determine a first standard vibration feature set, wherein the first standard vibration feature set includes a first standard abnormal frequency component, a first standard abnormal frequency ratio, a first standard amplitude curve, and a first standard waveform curve; Establishing a first mapping association between the first standard vibration feature set and the first standard defect type and the first defect intensity interval, and constructing a first defect recognition branch according to the first mapping association; A plurality of defect recognition branches are analyzed and acquired in sequence, and the defect feature recognition library is constructed according to the plurality of defect recognition branches.

6. The method for detecting defects in motor magnetic steel sheets based on vibration analysis according to claim 5, characterized in that: Inputting the multiple vibration feature sets into the defect feature recognition library for similarity traversal comparison respectively, selecting defect features whose similarity meets the dynamic comparison threshold, and outputting the multiple defect feature sets, including: Randomly selecting a first vibration feature set, and randomly selecting a first defect recognition branch in the defect feature recognition library; Input the first vibration feature set into the first defect recognition branch, perform similarity traversal comparison with the first standard vibration feature set, and if the comparison similarity satisfies the dynamic comparison threshold, output the first standard defect type and the first defect intensity interval as the first defect feature; The first vibration feature set is sequentially compared with standard vibration feature sets of other defect recognition branches in the defect feature recognition library through similarity traversal to obtain a first defect feature set, which is added to the multiple defect feature sets.

7. The method for detecting defects of motor magnetic steel sheets based on vibration analysis according to claim 6, characterized in that: Inputting the first vibration feature set into the first defect recognition branch, performing similarity traversal comparison with the first standard vibration feature set, and if the comparison similarity satisfies the dynamic comparison threshold, outputting the first standard defect type and the first defect intensity interval as the first defect feature, including: Acquire the first vibration feature set, wherein the first vibration feature set includes a first abnormal frequency component, a first abnormal frequency ratio, a first amplitude curve, and a first waveform curve; performing a similarity comparison between the first abnormal frequency component and a first standard abnormal frequency component to determine a first similarity; If the first similarity is greater than the dynamic comparison threshold, performing a similarity comparison on the first abnormal frequency ratio and the first standard abnormal frequency ratio to determine a second similarity; If the second similarity is greater than the dynamic comparison threshold, performing a similarity comparison on the first amplitude curve and the first standard amplitude curve to determine a third similarity; If the third similarity is greater than the dynamic comparison threshold, a similarity comparison is performed on the first waveform curve and the first standard waveform curve to determine a fourth similarity. If the fourth similarity is greater than the dynamic comparison threshold, the output first standard defect type and the first defect intensity range are set as the first defect feature.

8. The method for detecting defects in motor magnetic steel sheets based on vibration analysis according to claim 5, characterized in that: The plurality of initial defect features are analyzed for credible probability according to the temperature distribution feature and the acoustic wave feature, and the initial defect feature with the highest probability is output as the motor magnetic steel sheet defect detection result of the target motor, including: Taking the motor property characteristics and the motor magnetic steel sheet property characteristics as equipment constraints, taking the defect type and defect intensity as feature constraints, and taking the motor magnetic steel sheet defect detection as a guide, retrieve and obtain a sample temperature distribution feature set, a sample acoustic wave feature set, and multiple sample defect feature sets, and count the proportions of sample defect features under different sample temperature distribution characteristics and different sample acoustic wave characteristics, set them as the trigger probability of multiple defect features, construct a sample defect feature probability distribution, and obtain a sample defect feature probability distribution set; Performing supervised training on N prediction operators by using the sample temperature distribution feature set, the sample acoustic wave feature set, multiple sample defect feature sets and the sample defect feature probability distribution set until convergence, obtaining N defect feature trigger probability prediction units, and constructing a defect feature trigger probability predictor based on the N defect feature trigger probability prediction units, wherein the prediction operators at least include a BP neural network, a random forest and a support vector machine, and N is an integer greater than or equal to 3; The temperature distribution characteristics, acoustic wave characteristics and multiple initial defect characteristics are input into the defect characteristic trigger probability predictor for credible probability analysis, and the defect characteristic probability distribution is output. The initial defect characteristic with the maximum probability in the defect characteristic probability distribution is selected as the motor magnetic steel sheet defect detection result of the target motor.

Images