Axial magnetic field motor rotor assembly fault detection method and system

Abstract

The invention relates to an axial magnetic field motor rotor assembly fault detection method and system, and the method comprises the steps: collecting vibration waveform data, temperature gradient distribution data and magnetic field intensity distribution data of a rotor assembly in an operation state, and constructing a dynamic monitoring data set; performing dynamic feature extraction on the data in the dynamic monitoring data set to generate a dynamic operation feature set comprising a vibration coupling feature, a temperature abnormal fluctuation feature and a magnetic field offset feature, and performing multi-dimensional correlation mapping processing based on the dynamic operation feature set to generate a comprehensive abnormal index set; according to a matching result between the comprehensive abnormal index set and a preset fault judgment threshold value, determining a fault type set of the target rotor assembly and a corresponding fault confidence score; according to the fault type set of the rotor assembly and the corresponding fault confidence score, a visual detection report is generated, the visual detection report is used for displaying a fault result at a terminal, and accurate detection and visual presentation of the fault of the rotor assembly are achieved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular, to a method and system for detecting faults in a rotor assembly of an axial magnetic field motor. Background Art

[0002] Axial magnetic field motors have been widely used due to their unique structural and performance advantages, such as high power density and compact axial dimensions. However, during long-term operation, the rotor assembly of an axial magnetic field motor is prone to various faults due to the combined effects of mechanical stress, electromagnetic interference, thermal effects, and other factors.

[0003] Existing methods for detecting faults in the rotor assembly of an axial magnetic field motor have certain limitations. Some methods rely only on monitoring data in a single dimension, such as only monitoring vibration data or temperature data. This approach is difficult to comprehensively and accurately reflect the operating state of the rotor assembly, and is prone to missing or misjudging faults. In addition, when dealing with multi-dimensional data, some detection methods lack effective data fusion and correlation analysis means, and cannot fully explore the internal relationships between data in different dimensions, resulting in inaccurate judgment of fault types and difficulty in accurately predicting the occurrence of faults in advance, posing a potential threat to the safe and stable operation of the equipment. Summary of the Invention

[0004] The object of the present invention is to provide a method and system for detecting faults in a rotor assembly of an axial magnetic field motor, aiming to more comprehensively, accurately, and efficiently detect faults in the rotor assembly of an axial magnetic field motor.

[0005] To achieve the above object, in the first aspect of the embodiments of the present disclosure, a method for detecting faults in a rotor assembly of an axial magnetic field motor is provided. The method includes: Collect vibration waveform data, temperature gradient distribution data, and magnetic field intensity distribution data of the rotor assembly in the axial magnetic field motor during operation, and construct a dynamic monitoring data set according to the vibration waveform data, the temperature gradient distribution data, and the magnetic field intensity distribution data; Perform dynamic feature extraction processing on the data in the dynamic monitoring data set to generate a dynamic operation feature set including vibration coupling features, temperature abnormal fluctuation features, and magnetic field offset features, and perform multi-dimensional correlation mapping processing based on the dynamic operation feature set to generate a comprehensive abnormal index set of the rotor assembly; Determine the fault type set and the corresponding fault confidence score of the target rotor assembly with faults according to the matching result between the comprehensive abnormal index set and a preset fault determination threshold; Generate a visual inspection report for the target rotor assembly according to the set of fault types and corresponding fault confidence scores of the target rotor assembly, where the visual inspection report is used to display the fault results on a terminal.

[0006] In some possible implementation manners, the dynamic feature extraction process for the data in the dynamic monitoring data set to generate a dynamic operation feature set including vibration coupling features, temperature abnormal fluctuation features, and magnetic field offset features includes: Perform a frequency-domain transformation process on the vibration waveform data to generate the vibration coupling features; Perform a sliding window statistical analysis on the temperature gradient distribution data, calculate the slope of the temperature change trend corresponding to each temperature sensing unit, and use the difference value between the slopes of the temperature change trends of adjacent temperature sensing units as the temperature abnormal fluctuation feature; Perform a spatial distribution fitting on the magnetic field intensity distribution data to generate an axial magnetic field intensity gradient curve, and determine the magnetic field offset feature according to the curvature change points of the gradient curve; Perform a sliding window normalization process on the vibration coupling features, temperature abnormal fluctuation features, and magnetic field offset features respectively, and correlate the normalized features in the time-space dimension to generate the dynamic operation feature set including time sequence and spatial correlation.

[0007] In some possible implementation manners, the performing a frequency-domain transformation process on the vibration waveform data to generate the vibration coupling features includes: Divide the vibration waveform data into multiple equal-length time sequence segments, perform a fast Fourier transform process on each time sequence segment to generate a corresponding vibration spectrum segment; Extract the fundamental frequency component amplitude, harmonic component amplitude, and high-frequency noise energy value of each vibration spectrum segment; Perform a sliding average process on the fundamental frequency component amplitudes, harmonic component amplitudes, and high-frequency noise energy values of multiple time sequence segments to generate continuous vibration spectrum features; Set a first target frequency band corresponding to the resonance frequency of the rotor structure of the rotor assembly, and calculate a first proportional coefficient of the integral value of the vibration energy in the first target frequency band to the integral value of the vibration energy in the entire frequency band; Set a second target frequency band corresponding to the bearing wear characteristic frequency of the rotor assembly, and calculate a second proportional coefficient of the integral value of the vibration energy in the second target frequency band to the integral value of the vibration energy in the entire frequency band; Set the weight distribution ratio of the first proportional coefficient and the second proportional coefficient according to the historical fault data correlation between the rotor structure resonance frequency and the bearing wear characteristic frequency; Take the sum of the weighted first proportional coefficient and the second proportional coefficient as the vibration coupling feature.

[0008] In some possible implementation manners, the sliding window statistical analysis of the temperature gradient distribution data, calculating the temperature change trend slope corresponding to each temperature sensing unit, and taking the difference value between the temperature change trend slopes of adjacent temperature sensing units as the temperature abnormal fluctuation feature includes: Set the length of the sliding window to a preset time period, perform linear regression analysis on the temperature data of each temperature sensing unit within the sliding window to obtain a temperature change trend equation; Extract the slope parameter in the temperature change trend equation as the temperature change trend slope corresponding to this temperature sensing unit; When the temperature change trend slope exceeds a preset positive change threshold, mark this temperature sensing unit as an abnormal temperature rise unit; Obtain the physical distance parameter between adjacent temperature sensing units, and calculate the temperature conduction delay compensation coefficient based on the physical distance parameter; Perform time alignment processing on the temperature change trend slopes of adjacent temperature sensing units according to the temperature conduction delay compensation coefficient; Calculate the absolute value of the slope difference between adjacent temperature sensing units after alignment, and compare the absolute value of the difference with a preset difference threshold to generate the temperature abnormal fluctuation feature.

[0009] In some possible implementation manners, the spatial distribution fitting of the magnetic field intensity distribution data to generate an axial magnetic field intensity gradient curve includes: Construct a three-dimensional magnetic field intensity distribution surface according to the spatial coordinate positions of multiple magnetic field detection devices; Perform slicing processing on the three-dimensional distribution surface along the axial direction of the rotor assembly, and extract the magnetic field intensity distribution data of each slice plane; Perform polynomial fitting on the magnetic field intensity distribution data of each slice plane to generate a corresponding axial magnetic field intensity gradient curve; Calculate the similarity coefficient between the gradient curves of adjacent slice planes, and mark the slice plane with a similarity coefficient lower than the preset threshold as a magnetic field abnormal area.

[0010] In some possible implementation manners, the acquisition of the vibration waveform data, temperature gradient distribution data, and magnetic field intensity distribution data of the rotor assembly in the axial magnetic field motor during the operating state, and constructing a dynamic monitoring data set according to the vibration waveform data, the temperature gradient distribution data, and the magnetic field intensity distribution data includes: Deploy a plurality of vibration sensors at preset positions of the rotor assembly, and collect the vibration waveform data through the vibration sensors with a first sampling period. The vibration waveform data includes the time-series waveforms of the axial vibration component and the radial vibration component; Set a plurality of temperature sensing units on the surface of the rotor assembly, and collect the temperature gradient distribution data through the temperature sensing units with a second sampling period. The temperature gradient distribution data includes the temperature change rate of different regions and the temperature difference value between adjacent regions; Arrange a plurality of magnetic field detection devices along the circumferential direction of the rotor assembly, and collect the magnetic field intensity distribution data through the magnetic field detection devices with a third sampling period. The magnetic field intensity distribution data includes the axial magnetic field intensity distribution curve and the magnetic field asymmetry coefficient; Perform timestamp synchronization processing on the vibration waveform data, temperature gradient distribution data, and magnetic field intensity distribution data respectively, and use interpolation or resampling methods to unify the data with inconsistent sampling frequencies to the same time reference; Establish a data mapping table according to the spatial position relationship of the sensors, and fuse the synchronized data in a time-space alignment manner to generate the dynamic monitoring data set.

[0011] In some possible implementation manners, perform multi-dimensional correlation mapping processing based on the dynamic operation feature set to generate the comprehensive abnormal index set of the rotor assembly, including: Establish a first correlation mapping relationship between the vibration coupling feature and the temperature abnormal fluctuation feature, and calculate the vibration-temperature coupling abnormal index based on the first correlation mapping relationship; Establish a second correlation mapping relationship between the temperature abnormal fluctuation feature and the magnetic field offset feature, and calculate the temperature-magnetic field interaction abnormal index based on the second correlation mapping relationship; Establish a third correlation mapping relationship between the magnetic field offset feature and the vibration coupling feature, and calculate the magnetic field-vibration synchronization abnormal index based on the third correlation mapping relationship; Perform normalization processing on the vibration-temperature coupling abnormal index, temperature-magnetic field interaction abnormal index, and magnetic field-vibration synchronization abnormal index respectively, allocate dynamic weight coefficients according to the contribution degrees of various abnormal indexes in the historical fault data, and perform weighted fusion on the normalized abnormal indexes based on the dynamic weight coefficients to generate the comprehensive abnormal index set.

[0012] In some possible implementation manners, the determining the fault type set and the corresponding fault confidence score of the target rotor assembly according to the matching result between the comprehensive abnormal index set and the preset fault determination threshold includes: Match the vibration-temperature coupling anomaly index with the first fault threshold range. If the matching result is beyond the first fault threshold range, determine that there is a rotor imbalance fault type, and calculate the first confidence score based on the exceeding amplitude; Match the temperature-magnetic field interaction anomaly index with the second fault threshold range. If the matching result is beyond the second fault threshold range, determine that there is a bearing overheating fault type, and calculate the second confidence score based on the exceeding amplitude; Match the magnetic field-vibration synchronization anomaly index with the third fault threshold range. If the matching result is beyond the third fault threshold range, determine that there is a magnetic pole offset fault type, and calculate the third confidence score based on the exceeding amplitude; Combine the rotor imbalance fault type, the bearing overheating fault type, and the magnetic pole offset fault type to generate the fault type set of the target rotor assembly with faults, and combine the first confidence score, the second confidence score, and the third confidence score to generate the fault confidence score of the target rotor assembly with faults.

[0013] In a second aspect of the embodiments of the present disclosure, there is provided a fault detection system for an axial magnetic field motor rotor assembly. The system includes: A construction module configured to collect vibration waveform data, temperature gradient distribution data, and magnetic field intensity distribution data of the rotor assembly in the axial magnetic field motor during the operating state, and construct a dynamic monitoring data set according to the vibration waveform data, the temperature gradient distribution data, and the magnetic field intensity distribution data; A first generation module configured to perform dynamic feature extraction processing on the data in the dynamic monitoring data set to generate a dynamic operation feature set including vibration coupling features, temperature anomaly fluctuation features, and magnetic field offset features, and perform multi-dimensional correlation mapping processing based on the dynamic operation feature set to generate a comprehensive anomaly index set of the rotor assembly; A determination module configured to determine the fault type set and the corresponding fault confidence score of the target rotor assembly with faults according to the matching result between the comprehensive anomaly index set and a preset fault determination threshold; A second generation module configured to generate a visual detection report for the target rotor assembly according to the fault type set and the corresponding fault confidence score of the target rotor assembly, and the visual detection report is used to display the fault result on the terminal.

[0014] In a third aspect of the embodiments of the present disclosure, there is provided an electronic device, including: A memory on which a computer program is stored; A processor configured to execute the computer program in the memory to implement the steps of the method according to any one of the first aspects.

[0015] The present invention provides a method and system for fault detection of an axial magnetic field motor rotor assembly. Compared with the prior art, the following beneficial effects are achieved: By collecting a multi-dimensional dynamic monitoring data set, which comprehensively covers information such as vibration, temperature, and magnetic field strength, it can reflect the operating state of the target rotor assembly more completely compared with single-dimensional monitoring data. Performing dynamic feature extraction and multi-dimensional correlation mapping processing on the multi-dimensional data, the internal relationships between data in different dimensions are mined, and the generated comprehensive anomaly index set can more accurately reflect the abnormal conditions of the rotor assembly. Determining the fault type set and the corresponding fault confidence score based on the matching result between the comprehensive anomaly index set and the preset fault determination threshold improves the accuracy and reliability of fault judgment. Outputting a visual detection report and transmitting it to the target terminal for display, which facilitates relevant personnel to intuitively and timely grasp the fault situation of the rotor assembly, so that corresponding maintenance measures can be quickly taken to ensure the safe and stable operation of the axial magnetic field motor and reduce the maintenance cost and losses caused by faults of the axial magnetic field motor.

[0016] Other features and advantages of the present disclosure will be described in detail in the following specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following specific implementation, they are used to explain the present disclosure, but do not constitute a limitation to the present disclosure. In the drawings: Figure 1 is a flowchart of a method for fault detection of an axial magnetic field motor rotor assembly shown according to an embodiment of the specification.

[0018] Figure 2 is a block diagram of a system for fault detection of an axial magnetic field motor rotor assembly shown according to an embodiment of the specification.

[0019] Figure 3 is a block diagram of a device for executing a method for fault detection of an axial magnetic field motor rotor assembly shown according to an embodiment of the specification. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0021] The following will describe the specific embodiments of the present disclosure in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for explaining and understanding the present disclosure, and are not used to limit the present disclosure.

[0022] The present disclosure provides a method for detecting faults in a rotor assembly of an axial magnetic field motor, which is applied to a motor controller. Figure 1 FIG. 5 is a flowchart of a method for detecting faults in a rotor assembly of an axial magnetic field motor according to an embodiment. Specifically, the method includes: In step S11, vibration waveform data, temperature gradient distribution data, and magnetic field intensity distribution data of the rotor assembly in the axial magnetic field motor during the operating state are collected, and a dynamic monitoring data set is constructed based on the vibration waveform data, the temperature gradient distribution data, and the magnetic field intensity distribution data. Among them, the axial magnetic field motor can be a motor applied to, for example, an AGV autonomous vehicle. The axial magnetic field motor has the advantages of a compact structure, a high power density, and high efficiency, and can provide strong and efficient power support for the autonomous vehicle. The rotor assembly is the rotating part in the axial magnetic field motor and is usually composed of a rotor core, a rotor winding, etc. When the AGV autonomous vehicle is running, the state of the rotor assembly directly affects the performance of the motor and the driving stability of the autonomous vehicle.

[0023] Among them, the vibration waveform data can be collected by vibration sensors installed on or near the rotor assembly, and is waveform data reflecting the change of the vibration condition of the rotor assembly over time during the operation process. It contains rich information on the mechanical motion state of the rotor assembly and can be used to judge whether there are mechanical faults.

[0024] The temperature gradient distribution data is data used to describe the temperature difference at different positions of the rotor assembly. It is obtained by arranging temperature sensors at different parts of the rotor assembly, and can reflect the heat conduction condition and local overheating phenomenon of the rotor assembly, which helps to detect faults such as winding short circuits and insulation damage.

[0025] The magnetic field intensity distribution data is used to represent the spatial distribution of the magnetic field intensity around the rotor assembly. It is measured by magnetic field sensors and can reflect the electromagnetic performance of the motor and judge whether there are problems such as magnetic field distortion and pole demagnetization.

[0026] In the embodiment of the present disclosure, during the operation of the AGV autonomous vehicle, the rotor assembly of the axial magnetic field motor will generate vibrations, temperature changes, and magnetic field changes. By reasonably arranging vibration sensors, temperature sensors, and magnetic field sensors on and around the rotor assembly, vibration waveform data, temperature gradient distribution data, and magnetic field intensity distribution data are collected in real time. These data are sorted and stored according to the time sequence and spatial position to construct a dynamic monitoring data set for subsequent analysis and diagnosis of the operating state of the rotor assembly.

[0027] For example, on the axial magnetic field motor of an AGV driverless vehicle, three vibration sensors are installed at different positions (such as the top, middle, and bottom) of the rotor assembly, two temperature sensors are used to measure the temperatures of the rotor winding and the iron core respectively, and one magnetic field sensor measures the magnetic field intensity around the rotor. After the AGV driverless vehicle starts and runs for a period of time, the vibration sensors collect vibration waveform data at different positions, the temperature sensors record the temperature change data of the winding and the iron core, and the magnetic field sensor obtains the magnetic field intensity distribution data. Integrating these data together constitutes the dynamic monitoring data set of the rotor assembly during this period.

[0028] In step S12, the data in the dynamic monitoring data set is subjected to dynamic feature extraction processing to generate a dynamic operation feature set including vibration coupling features, temperature abnormal fluctuation features, and magnetic field offset features, and based on the dynamic operation feature set, multi-dimensional correlation mapping processing is performed to generate the comprehensive abnormal index set of the rotor assembly; Among them, the vibration coupling feature is the feature manifested by the mutual coupling and mutual influence between different frequency components in the vibration signal. In the axial magnetic field motor of an AGV driverless vehicle, the vibration coupling feature can reflect the interaction relationship between multiple mechanical components (such as bearings, rotor iron cores, etc.) in the rotor assembly, which helps to discover complex mechanical faults.

[0029] The temperature abnormal fluctuation feature is the abnormal change feature that appears in the temperature data, such as the temperature suddenly rising, falling, or having too large a fluctuation amplitude, etc. During the operation of the AGV driverless vehicle, the temperature abnormal fluctuation feature may indicate problems such as local overheating, poor heat dissipation, or electrical faults in the rotor assembly.

[0030] The magnetic field offset feature is the deviation phenomenon that appears in the magnetic field intensity distribution data compared with the normal state in terms of the magnetic field intensity. In the axial magnetic field motor, the magnetic field offset feature may mean faults such as abnormal pole positions, magnetic circuit blockages, or pole demagnetization.

[0031] In the embodiments of the present disclosure, for the data in the dynamic monitoring data set, dynamic feature extraction processing is performed. Signal processing techniques (such as Fourier transform, wavelet analysis, etc.) are used to extract vibration coupling features from the vibration waveform data, identify temperature abnormal fluctuation features from the temperature gradient distribution data, and detect magnetic field offset features from the magnetic field intensity distribution data. Then, based on the extracted dynamic operation feature set, data mining and machine learning algorithms (such as association rule mining, neural networks, etc.) are used for multi-dimensional correlation mapping processing to analyze the mutual relationship between different features and generate a comprehensive abnormal index set that can comprehensively reflect the operation state of the rotor assembly.

[0032] For example, for the collected vibration waveform data, it is transformed from the time domain to the frequency domain through Fourier transform, the amplitude and phase relationships of different frequency components are analyzed, and vibration coupling characteristics are extracted, such as the harmonic relationship or coupling frequency between certain frequency components.

[0033] For the temperature gradient distribution data, a temperature change threshold is set. When the temperature change exceeds this threshold, it is determined as a characteristic of abnormal temperature fluctuation. For the magnetic field intensity distribution data, the deviation between the magnetic field intensity at different positions and the normal value is calculated to obtain the magnetic field offset characteristic. Then, the association rule mining algorithm is used to analyze the association relationships among the vibration coupling characteristic, the abnormal temperature fluctuation characteristic, and the magnetic field offset characteristic. For example, it is found that when a harmonic wave of a specific frequency appears in the vibration coupling characteristic and the abnormal temperature fluctuation characteristic shows a sudden increase in temperature, the magnetic field offset characteristic will also increase accordingly. According to these association relationships, a set of comprehensive anomaly indicators is generated, such as a comprehensive anomaly index, a fault risk level, etc.

[0034] In step S13, according to the matching result between the set of comprehensive anomaly indicators and a preset fault determination threshold, the set of fault types and the corresponding fault confidence scores of the target rotor assembly with a fault are determined; In the embodiment of the present disclosure, the generated set of comprehensive anomaly indicators is compared and matched with a preset fault determination threshold. According to the matching result, it is judged whether the rotor assembly has a fault and the possible fault types. For each possible fault type, according to its correlation and severity with the comprehensive anomaly indicators, a corresponding fault confidence score is given. The fault confidence score can be determined by methods such as expert experience, historical data statistics, or machine learning models.

[0035] For example, it is assumed that in the preset fault determination threshold, when the comprehensive anomaly index is greater than 80, it is considered that there is a fault. In the generated set of comprehensive anomaly indicators, the comprehensive anomaly index is 85, which exceeds the fault determination threshold. Therefore, it is determined that the rotor assembly has a fault. Further analyzing the association between the comprehensive anomaly indicators and different fault types, it is found that when the comprehensive anomaly index is between 80 and 90, the possibility of bearing failure is relatively high, and the possibility of winding short - circuit is the second. According to historical data statistics, in this case, the confidence score for bearing failure is 0.8, and the confidence score for winding short - circuit is 0.3. Therefore, the set of fault types of the rotor assembly is determined as {bearing failure, winding short - circuit}, and the corresponding fault confidence scores are {0.8, 0.3}.

[0036] In step S14, according to the set of fault types and the corresponding fault confidence scores of the target rotor assembly, a visual inspection report for the target rotor assembly is generated, and the visual inspection report is used to display the fault result on the terminal.

[0037] Among them, the visual inspection report is a report used to present the set of fault types of the rotor assembly and the corresponding fault confidence scores in an intuitive and visual manner. The report may contain elements such as charts and graphs, facilitating the display and analysis of fault results by users at the terminal.

[0038] In the embodiments of the present disclosure, according to the determined set of fault types of the rotor assembly and the corresponding fault confidence scores, a visual inspection report is generated using visualization techniques (such as charts, graphs, reports, etc.). The report may include a bar chart showing the confidence scores of different fault types, a line chart showing the change trend of the comprehensive anomaly index over time, as well as detailed fault descriptions and recommended measures. The generated visual inspection report is transmitted to the terminal device, facilitating users to display and analyze the fault results.

[0039] For example, the Matplotlib library in Python is used to generate a visual inspection report. The report contains a bar chart, where the abscissa represents the fault types (bearing fault, winding short circuit), and the ordinate represents the fault confidence scores. The bar chart intuitively shows that the confidence score of the bearing fault is 0.8, and the confidence score of the winding short circuit is 0.3.

[0040] At the same time, there is also a line chart in the report, showing the change trend of the comprehensive anomaly index over a period of time. Users can understand the change situation of the operating state of the rotor assembly through the line chart. In addition, the report also contains a detailed description of the fault, such as the possible causes of the bearing fault (bearing wear, insufficient lubrication, etc.) and recommended measures (replace the bearing, strengthen lubrication, etc.). The report is transmitted to the monitoring terminal of the AGV autonomous vehicle, and the operator can clearly see the fault information on the terminal, so as to take corresponding measures in a timely manner.

[0041] The above technical solution collects a multi-dimensional dynamic monitoring data set, comprehensively covering information such as vibration, temperature, and magnetic field intensity. Compared with single-dimensional monitoring data, it can more completely reflect the operating state of the target rotor assembly. By performing dynamic feature extraction and multi-dimensional correlation mapping processing on the multi-dimensional data, the internal connections between different-dimensional data are mined, and the generated comprehensive anomaly index set can more accurately reflect the abnormal conditions of the rotor assembly. Based on the matching results of the comprehensive anomaly index set and the preset fault determination threshold, the set of fault types and the corresponding fault confidence scores are determined, improving the accuracy and reliability of fault judgment. The visual inspection report is output and transmitted to the target terminal for display, facilitating relevant personnel to intuitively and timely grasp the fault situation of the rotor assembly, so that corresponding maintenance measures can be taken quickly, ensuring the safe and stable operation of the axial magnetic field motor, and reducing the maintenance cost of the axial magnetic field motor and the losses caused by faults.

[0042] In a possible implementation manner, in step S12, the process of performing dynamic feature extraction on the data in the dynamic monitoring data set to generate a dynamic operation feature set including vibration coupling features, temperature abnormal fluctuation features, and magnetic field offset features includes: In step S121, perform frequency domain transformation processing on the vibration waveform data to generate the vibration coupling features; In the embodiments of the present disclosure, the vibration waveform data is a time-domain signal that changes with time, and it is difficult to directly extract features reflecting the mutual relationship of different frequency components from it. Through frequency domain transformation (such as Fourier transformation), the time-domain vibration signal is converted to the frequency domain to obtain the amplitude and phase information of the signal at different frequencies. In the frequency domain, the coupling relationships between different frequency components can be analyzed, such as harmonic coupling, sideband coupling, etc. These coupling relationships reflect the mechanical interactions between various components (such as bearings, rotor cores, etc.) in the rotor assembly, thereby generating vibration coupling features.

[0043] For example, install a vibration sensor on the axial magnetic field motor of an AGV autonomous vehicle to collect the vibration waveform data of the rotor assembly during operation. Use the fast Fourier transform (FFT) to convert the time-domain vibration signal into a frequency-domain signal. Analyzing the frequency-domain signal reveals that additional amplitude peaks appear at integer multiples of the fundamental frequency (such as the rotation frequency of the motor), and there is a certain phase relationship between these peaks, indicating the presence of harmonic coupling. At the same time, some sidebands related to the fundamental frequency are also found, and the frequency interval of the sidebands is related to the fault characteristic frequency of the bearing, which further reflects the mechanical coupling relationship between the bearing and the rotor. These features such as harmonic coupling and sidebands are used as vibration coupling features.

[0044] In step S122, perform sliding window statistical analysis on the temperature gradient distribution data, calculate the temperature change trend slope corresponding to each temperature sensing unit, and use the difference value between the temperature change trend slopes of adjacent temperature sensing units as the temperature abnormal fluctuation feature; Among them, sliding window statistical analysis can be performed by setting a window with a fixed size to slide on the data sequence and performing statistical analysis on the data within the window to capture the local change trend of the data.

[0045] In the embodiments of the present disclosure, the temperature gradient distribution data reflects the temperature change conditions at different positions of the motor rotor assembly. Through sliding window statistical analysis, linear fitting is performed on the temperature data within each sliding window, and the slope of the temperature change trend over time within the window is calculated, that is, the temperature change rate. The slopes of the temperature change trends of adjacent temperature sensing units should have a certain similarity. If the difference value between adjacent slopes is too large, it indicates that there are abnormal fluctuations in the temperature change in this area, which may be caused by reasons such as local overheating, poor heat dissipation, or electrical faults. This difference value is used as the temperature abnormal fluctuation feature.

[0046] For example, a plurality of temperature sensors are evenly arranged on the axial magnetic field motor rotor of the AGV driverless vehicle to collect temperature gradient distribution data. The size of the sliding window is set to 10 data points, and the window slides 1 data point each time. For each temperature sensing unit, least squares method is used for linear fitting within each sliding window to obtain the slope of the temperature change trend. For example, at a certain moment, the slope of the temperature change trend within the current sliding window of temperature sensing unit A is 0.5 °C / s, and the slope of the adjacent temperature sensing unit B is -0.2 °C / s. The difference value between the two is 0.7 °C / s. If this difference value exceeds the preset threshold (such as 0.3 °C / s), it is considered that there is an abnormal temperature fluctuation in this area, and this difference value is used as part of the temperature abnormal fluctuation feature.

[0047] In step S123, spatial distribution fitting is performed on the magnetic field intensity distribution data to generate an axial magnetic field intensity gradient curve, and the magnetic field offset feature is determined according to the curvature change points of the gradient curve; Among them, spatial distribution fitting can use mathematical methods to fit the magnetic field intensity distribution data to obtain a continuous distribution model of the magnetic field intensity in space, which is convenient for analyzing the characteristics of the magnetic field. The axial magnetic field intensity gradient curve is a curve generated by performing spatial distribution fitting on the magnetic field intensity distribution data, which describes the change trend of the magnetic field intensity along the axis. The curvature change points are the points where the curvature changes significantly in the axial magnetic field intensity gradient curve. These points may correspond to abnormal situations in the magnetic field distribution and are used to determine the magnetic field offset feature.

[0048] In the embodiments of the present disclosure, the magnetic field intensity distribution data are discrete spatial data points. Through a spatial distribution fitting method (such as polynomial fitting, spline interpolation, etc.), these discrete data points are fitted into a continuous magnetic field intensity distribution model. Then, the derivative of the fitted magnetic field intensity distribution along the axial direction is obtained to get the axial magnetic field intensity gradient curve, which describes the variation trend of the magnetic field intensity along the axial direction. Under normal circumstances, the magnetic field intensity gradient curve should be smooth. When there are magnetic circuit faults in the motor (such as pole demagnetization, magnetic circuit blockage, etc.), the magnetic field intensity distribution will change abnormally, resulting in curvature change points in the gradient curve. By detecting these curvature change points, the magnetic field offset characteristics can be determined.

[0049] For example, magnetic field sensors are arranged around the axial magnetic field motor of an AGV autonomous vehicle to collect magnetic field intensity distribution data. The quadratic polynomial fitting method is used to fit the magnetic field intensity distribution data to obtain a continuous distribution function of the magnetic field intensity in space. The derivative of this function along the axial direction is taken to generate the axial magnetic field intensity gradient curve. Analyzing the gradient curve reveals that at a certain axial position, the curvature of the curve changes significantly, and an inflection point appears on the originally smooth curve. The axial position corresponding to this inflection point may indicate an offset in the magnetic field distribution at that location, possibly due to the weakening of the magnetism of a certain pole or the presence of foreign objects in the magnetic circuit. The position and related characteristics of this curvature change point are used as the magnetic field offset characteristics.

[0050] In step S124, sliding window normalization processing is respectively performed on the vibration coupling characteristics, temperature abnormal fluctuation characteristics, and magnetic field offset characteristics, and the normalized characteristics are correlated according to the time - space dimension to generate the dynamic operation characteristic set including time - series and space correlation.

[0051] Among them, sliding window normalization processing is to perform normalization processing on the extracted characteristic data to make them have the same dimension and distribution characteristics, which is convenient for subsequent comparison and analysis. Sliding window normalization is to perform normalization operations on the data within a sliding window.

[0052] In the embodiments of the present disclosure, since the vibration coupling characteristics, temperature abnormal fluctuation characteristics, and magnetic field offset characteristics may have different dimensions and distribution characteristics, direct comparison and analysis may be affected by the dimension differences. Therefore, sliding window normalization processing is adopted to normalize the data of each characteristic within the sliding window (such as Z - score normalization) to make them have the same mean and standard deviation. Then, these normalized characteristics are correlated according to the time and space dimensions. The time - dimension correlation can reflect the variation trend of the characteristics over time, and the space - dimension correlation can reflect the distribution of the characteristics at different positions, thereby generating a dynamic operation characteristic set including time - series and space correlation, providing more comprehensive information for subsequent fault diagnosis.

[0053] For example, for the extracted vibration coupling features, temperature anomaly fluctuation features, and magnetic field offset features, the sliding window size is set to 20 data points, and the window slides 5 data points each time. Within each sliding window, the Z-score normalization is performed on the vibration coupling features. The calculation formula is z = (x - μ) / σ, where x is the current feature value, μ is the mean value of the feature values within the window, and σ is the standard deviation of the feature values within the window. Similarly, the same normalization process is also performed on the temperature anomaly fluctuation features and magnetic field offset features.

[0054] In the time dimension, the normalized features are arranged in chronological order to analyze the changing trend of the features over time. For example, observe the fluctuation of the vibration coupling features over a period of time to determine whether there is a gradually increasing trend, which may indicate the gradual development of a fault. In the spatial dimension, the normalized features at different positions (such as different parts of the rotor) are correlated. For example, compare the temperature anomaly fluctuation features at adjacent positions. If the feature differences at adjacent positions are large, it may indicate a local fault in this area. Integrate the features after correlating the time and space dimensions to generate a dynamic operation feature set, such as a time series matrix containing multiple features, where each element corresponds to the feature value at a specific time and space position.

[0055] In a possible implementation manner, in step S121, the performing frequency domain transformation processing on the vibration waveform data to generate the vibration coupling features includes: Dividing the vibration waveform data into multiple equal-length time series segments, and performing fast Fourier transform processing on each time series segment to generate corresponding vibration spectrum segments; In the embodiments of the present disclosure, the collected vibration waveform data V is divided into multiple equal-length time series segments Vi, where i = 1, 2,..., n according to a fixed length L. Perform fast Fourier transform on each time series segment Vi to convert the vibration signal in the time domain to the frequency domain, and obtain the corresponding vibration spectrum segment Si. Through this transformation, the components of the vibration at different frequencies can be analyzed.

[0056] Extracting the fundamental frequency component amplitude, harmonic component amplitude, and high-frequency noise energy value of each vibration spectrum segment; In the embodiments of the present disclosure, from each vibration spectrum segment Si, find the fundamental frequency component amplitude Af, harmonic component amplitude Ahj (j represents different harmonic orders), and high-frequency noise energy value En. The fundamental frequency component amplitude reflects the intensity of the main frequency component of the vibration, the harmonic component amplitude reflects the vibration intensity at different harmonic frequencies, and the high-frequency noise energy value characterizes the influence of high-frequency noise on the vibration. For example, in Si, determine the amplitude Af corresponding to the fundamental frequency f, the amplitudes Ahj corresponding to each harmonic frequency fj, and calculate the energy value in the high-frequency region as En.

[0057] Perform a moving average process on the fundamental frequency component amplitudes, harmonic component amplitudes, and high-frequency noise energy values of multiple time series segments to generate continuous vibration spectrum features; In the embodiments of the present disclosure, a moving average process is performed on the fundamental frequency component amplitude Af, harmonic component amplitude Ahj, and high-frequency noise energy value En extracted from multiple time series segments. Set the moving window length to m. For the fundamental frequency component amplitude, calculate the moving average value Afa = (Afk + Afk-1 + … + Afk-m+1) / m, where k = m, m+1, …, n. Similarly, calculate the moving average values of the harmonic component amplitude and high-frequency noise energy value to generate the continuous vibration spectrum feature Sc, which more smoothly shows the variation of the vibration spectrum over time.

[0058] Set a first target frequency band corresponding to the resonance frequency of the rotor structure of the rotor assembly, and calculate a first proportionality coefficient of the integral value of the vibration energy in the first target frequency band to the integral value of the vibration energy in the entire frequency band; In the embodiments of the present disclosure, according to the structural parameters and material properties of the rotor, determine a first target frequency band Fr1 corresponding to the resonance frequency of the rotor structure. Calculate the integral value of the vibration energy Er1 in this frequency band, and the integral value of the vibration energy in the entire frequency band is Ea. The first proportionality coefficient Pr1 = Er1 / Ea, and this coefficient reflects the proportion of the vibration energy in the first target frequency band to the total vibration energy.

[0059] Set a second target frequency band corresponding to the bearing wear characteristic frequency of the rotor assembly, and calculate a second proportionality coefficient of the integral value of the vibration energy in the second target frequency band to the integral value of the vibration energy in the entire frequency band; In the embodiments of the present disclosure, based on the model, material, and operating conditions of the bearing, determine a second target frequency band Fr2 corresponding to the bearing wear characteristic frequency. Calculate the integral value of the vibration energy Er2 in this frequency band, and the second proportionality coefficient Pr2 = Er2 / Ea. This coefficient reflects the proportion of the vibration energy in the second target frequency band in the total vibration energy.

[0060] According to the historical fault data correlation between the rotor structure resonance frequency and the bearing wear characteristic frequency, set the weight distribution ratio of the first proportionality coefficient and the second proportionality coefficient; In the embodiments of the present disclosure, analyze a large amount of past rotor fault data to determine the correlation between the rotor structure resonance frequency, the bearing wear characteristic frequency, and various faults. For example, it is found that the probability ratio of the rotor structure resonance frequency causing faults is w1, and the probability ratio of the bearing wear characteristic frequency causing faults is w2, and w1 + w2 = 1. Based on this, set the weight of the first proportionality coefficient Pr1 to w1 and the weight of the second proportionality coefficient Pr2 to w2.

[0061] Take the sum of the weighted first proportionality coefficient and the second proportionality coefficient as the vibration coupling feature.

[0062] In the embodiments of the present disclosure, the vibration coupling feature Vc = w1×Pr1 + w2×Pr2, which comprehensively reflects the vibration coupling situation related to the resonance of the rotor structure and bearing wear.

[0063] In a possible implementation, in step S122, the sliding window statistical analysis of the temperature gradient distribution data, calculating the temperature change trend slope corresponding to each temperature sensing unit, and taking the difference value between the temperature change trend slopes of adjacent temperature sensing units as the temperature abnormal fluctuation feature, includes: Set the length of the sliding window to a preset time period, and perform linear regression analysis on the temperature data of each temperature sensing unit within the sliding window to obtain a temperature change trend equation; In the embodiments of the present disclosure, the length of the sliding window is set to Tw, and for the temperature data Tt collected by each temperature sensing unit, linear regression analysis is performed within the sliding window with a length of Tw. By methods such as the least squares method, a straight line equation T = at + b is found, where a and b are coefficients determined through regression analysis, and this straight line equation represents the temperature change trend of this temperature sensing unit within the sliding window.

[0064] Extract the slope parameter in the temperature change trend equation as the temperature change trend slope corresponding to this temperature sensing unit; In the embodiments of the present disclosure, from the temperature change trend equation T = at + b, the slope parameter a is extracted, and this slope a is the temperature change trend slope St corresponding to this temperature sensing unit, which reflects the rate of change of the temperature in the area where this temperature sensing unit is located over time.

[0065] When the temperature change trend slope exceeds a preset positive change threshold, mark this temperature sensing unit as an abnormal temperature rise unit; In the embodiments of the present disclosure, the preset positive change threshold is set to Th. If the temperature change trend slope St corresponding to a certain temperature sensing unit > Th, then mark this temperature sensing unit as an abnormal temperature rise unit, indicating that the temperature in this area rises too fast and there may be an abnormal situation.

[0066] Obtain the physical distance parameter between adjacent temperature sensing units, and calculate the temperature conduction delay compensation coefficient based on the physical distance parameter; In the embodiments of the present disclosure, the physical distance d between adjacent temperature sensing units is measured. According to the heat conduction theory and material properties, the temperature conduction delay compensation coefficient Ct is calculated. For example, Ct may be related to factors such as the distance d and the thermal conductivity of the material, and is calculated through relevant formulas to consider the delay effect of temperature conduction in adjacent areas.

[0067] Perform time alignment processing on the temperature change trend slopes of adjacent temperature sensing units according to the temperature conduction delay compensation coefficient; In the embodiments of the present disclosure, the temperature conduction delay compensation coefficient Ct is used to perform time alignment on the temperature change trend slopes St1 and St2 of adjacent temperature sensing units. For example, by adjusting the time coordinate or data processing method, the slope change caused by temperature conduction delay is synchronized in time to ensure the accuracy of subsequent calculation of the difference value.

[0068] Calculate the absolute value of the slope difference between adjacent temperature sensing units after alignment, and compare the absolute value of the difference with a preset difference threshold to generate the temperature abnormal fluctuation feature.

[0069] In the embodiments of the present disclosure, calculate the absolute value of the slope difference D = |St1 - St2| between adjacent temperature sensing units after alignment. Compare the absolute value of the difference D with a preset difference threshold Td. If D > Td, it indicates that the temperature change trend difference in the adjacent area is large and there is a temperature abnormal fluctuation situation, and thus generate the temperature abnormal fluctuation feature Ta.

[0070] In a possible implementation manner, in step S123, the spatial distribution fitting of the magnetic field intensity distribution data to generate an axial magnetic field intensity gradient curve includes: Construct a three-dimensional magnetic field intensity distribution surface according to the spatial coordinate positions of multiple magnetic field detection devices; In the embodiments of the present disclosure, according to the spatial coordinate positions (x, y, z) of multiple magnetic field detection devices in the circumferential and axial directions of the rotor assembly, and the magnetic field intensity values B collected by them, a three-dimensional magnetic field intensity distribution surface S(x, y, z) is constructed using a surface fitting algorithm. This surface comprehensively reflects the spatial distribution of the magnetic field.

[0071] Perform slicing processing on the three-dimensional distribution surface along the axial direction of the rotor assembly, and extract the magnetic field intensity distribution data of each slice plane; In the embodiments of the present disclosure, along the axial direction of the rotor assembly, the three-dimensional magnetic field intensity distribution surface S(x, y, z) is sliced at a certain interval. For example, it is sliced every Δz distance to obtain a series of slice planes. Extract the magnetic field intensity distribution data B(x, y) from each slice plane, and these data reflect the distribution of the magnetic field intensity on the plane at different axial positions.

[0072] Perform polynomial fitting on the magnetic field intensity distribution data of each slice plane to generate the corresponding axial magnetic field intensity gradient curve; In the embodiments of the present disclosure, for the magnetic field intensity distribution data B(x, y) of each slice plane, a polynomial fitting method is adopted, such as quadratic polynomial or cubic polynomial fitting. By fitting, a curve equation B = ax² + bx + c (taking the quadratic polynomial as an example) is obtained, and this curve is the corresponding axial magnetic field intensity gradient curve Bg(x), which shows the gradient change of the magnetic field intensity along a certain direction on this slice plane.

[0073] Calculate the similarity coefficient between the gradient curves of adjacent slice planes, and mark the slice plane with the similarity coefficient lower than the preset threshold as the magnetic field abnormal area.

[0074] In the embodiments of the present disclosure, through a specific similarity calculation method, such as calculating the Euclidean distance or correlation coefficient between two curves, etc., calculate the similarity coefficient Sc between the gradient curves Bg1(x) and Bg2(x) of adjacent slice planes. Set the preset threshold as Ts. If Sc < Ts, then mark this slice plane as the magnetic field abnormal area, indicating that the magnetic field distribution in this area is quite different from that of the adjacent area, and there may be abnormal situations such as magnetic field offset. According to the curvature change points of the gradient curve, for example, by calculating the curvature through derivation, find the points where the curvature changes significantly, and determine the magnetic field offset feature Mo.

[0075] In a possible implementation manner, in step S11, collect the vibration waveform data, temperature gradient distribution data, and magnetic field intensity distribution data of the rotor assembly in the axial magnetic field motor in the operating state, and construct a dynamic monitoring data set according to the vibration waveform data, the temperature gradient distribution data, and the magnetic field intensity distribution data, including: In step S111, deploy a plurality of vibration sensors at preset positions of the rotor assembly, and collect the vibration waveform data through the vibration sensors with a first sampling period. The vibration waveform data includes the time series waveforms of the axial vibration component and the radial vibration component. In the embodiments of the present disclosure, on the rotor assembly, according to its mechanical structure characteristics and past high-fault areas, select specific preset positions to arrange vibration sensors. For example, select several positions at equal intervals at both ends and in the middle of the rotor shaft, and install a plurality of high-precision vibration sensors. Collect the vibration waveform data with a first sampling period Ta, and Ta is determined according to the operating speed and vibration frequency characteristics of the motor. When the vibration sensor works, it records the waveform data of the axial vibration component Ax and the radial vibration component Ry changing with time t in real time. For example, at a certain moment t1, the vibration sensor records the axial vibration component Ax(t1) and the radial vibration component Ry(t1), and continuous collection forms continuous time series waveform data.

[0076] In step S112, a plurality of temperature sensing units are arranged on the surface of the rotor assembly, and the temperature gradient distribution data is collected through the temperature sensing units at a second sampling period. The temperature gradient distribution data includes the temperature change rate of different regions and the temperature difference value between adjacent regions; In the embodiments of the present disclosure, on the surface of the rotor assembly, a plurality of temperature sensing units are arranged according to a certain distribution rule. For example, they are arranged in a uniform matrix form to ensure coverage of key areas. Through these temperature sensing units, temperature data is collected at a second sampling period Tb. Tb needs to comprehensively consider the temperature change speed and the requirements for time resolution in subsequent data analysis. The temperature sensing units collect the temperature values T of each region, and thus the temperature change rate Tr=(T(t + Δt)-T(t)) / Δt of different regions can be calculated, where Δt is a set time interval. At the same time, for adjacent temperature sensing units, the temperature difference value Td = |T1 - T2| between adjacent regions can be calculated, where T1 and T2 are the temperature values collected by two adjacent temperature sensing units respectively. These data together constitute the temperature gradient distribution data.

[0077] In step S113, a plurality of magnetic field detection devices are arranged along the circumferential direction of the rotor assembly, and the magnetic field intensity distribution data is collected through the magnetic field detection devices at a third sampling period. The magnetic field intensity distribution data includes the axial magnetic field intensity distribution curve and the magnetic field asymmetry coefficient; In the embodiments of the present disclosure, along the circumferential direction of the rotor assembly, a plurality of magnetic field detection devices are arranged at equal angular intervals. These devices are used to detect the magnetic field information generated during the operation of the rotor. The magnetic field intensity distribution data is collected at a third sampling period Tc, and Tc is determined according to the fluctuation characteristics of the magnetic field and the measurement accuracy requirements. The data collected by the magnetic field detection devices can be processed to generate an axial magnetic field intensity distribution curve Az, which reflects the change of the magnetic field intensity along the axial position x. At the same time, the magnetic field asymmetry coefficient Ac is calculated through a specific algorithm, for example, by comparing the magnetic field intensity differences at different angular positions, etc., to measure the symmetry degree of the magnetic field in the circumferential direction.

[0078] In step S114, timestamp synchronization processing is respectively performed on the vibration waveform data, the temperature gradient distribution data, and the magnetic field intensity distribution data, and data with inconsistent sampling frequencies is unified to the same time reference by using interpolation or resampling methods; In the embodiments of the present disclosure, since the sampling frequencies of the vibration sensor, the temperature sensing unit, and the magnetic field detection device may be different, timestamps need to be added to the three types of data respectively to mark the exact time of data acquisition. If the sampling frequency Fa of the vibration waveform data is higher than the sampling frequency Fb of the temperature gradient distribution data, an interpolation method can be used. For example, using the linear interpolation method, appropriate data points are inserted within the time interval of the temperature gradient distribution data to make the time interval of the temperature data consistent with that of the vibration waveform data, so as to unify the three types of data to the same time reference.

[0079] In step S115, a data mapping table is established according to the spatial position relationship of the sensors, and the synchronized data is fused in a time-space alignment manner to generate the dynamic monitoring data set.

[0080] In the embodiments of the present disclosure, a data mapping table is constructed based on the spatial positions of the vibration sensor, the temperature sensing unit, and the magnetic field detection device on the rotor assembly. The specific position coordinates of each sensor are recorded. The data after timestamp synchronization processing is fused in a time-space alignment manner. That is, at the same moment t, the vibration waveform data, the temperature gradient distribution data, and the magnetic field intensity distribution data at the corresponding positions are integrated together to form a multi-dimensional dynamic monitoring data set M, and the dynamic monitoring data set can comprehensively reflect the operating state of the rotor assembly at moment t.

[0081] In a possible implementation manner, in step S12, the multi-dimensional correlation mapping process is performed based on the dynamic operation feature set to generate the comprehensive abnormal index set of the rotor assembly, including: In step S121, a first correlation mapping relationship is established between the vibration coupling feature and the temperature abnormal fluctuation feature, and the vibration-temperature coupling abnormal index is calculated based on the first correlation mapping relationship; In the embodiments of the present disclosure, the potential connection between the vibration coupling feature Vc and the temperature abnormal fluctuation feature Ta is analyzed, and a first correlation mapping relationship is established through methods such as historical data statistical analysis or theoretical derivation. For example, it is found that the change of the vibration coupling feature will affect the temperature abnormal fluctuation to a certain extent, and there may be a linear or non-linear relationship. Assuming a linear relationship, the vibration-temperature coupling abnormal index VTa = k1×Vc + k2×Ta, where k1 and k2 are coefficients determined by methods such as regression analysis of a large amount of historical fault data, and reflect the influence weights of the vibration coupling feature and the temperature abnormal fluctuation feature on this index.

[0082] In step S122, a second correlation mapping relationship is established between the temperature abnormal fluctuation feature and the magnetic field offset feature, and the temperature-magnetic field interaction abnormal index is calculated based on the second correlation mapping relationship; In the embodiments of the present disclosure, the relationship between the temperature anomaly fluctuation feature Ta and the magnetic field offset feature Mo is studied, and a second correlation mapping relationship is established through experimental data or theoretical models. For example, it may be found that there is a certain causal relationship between temperature anomaly fluctuations and magnetic field offsets. Assuming a non-linear relationship, the temperature-magnetic field interaction anomaly index TMo = f(Ta, Mo) is obtained through function fitting. The form of the function f is determined by analyzing historical data. For example, f may be a polynomial function containing Ta and Mo, and this index reflects the interaction anomaly between temperature and magnetic field.

[0083] In step S123, a third correlation mapping relationship between the magnetic field offset feature and the vibration coupling feature is established, and the magnetic field-vibration synchronization anomaly index is calculated based on the third correlation mapping relationship. In the embodiments of the present disclosure, the internal connection between the magnetic field offset feature Mo and the vibration coupling feature Vc is explored, and a third correlation mapping relationship is established with the help of data analysis tools and theoretical knowledge. For example, it is found that a magnetic field offset can cause a change in vibration characteristics, and vice versa. Assuming a complex non-linear relationship, a relationship is established through methods such as a neural network model. The magnetic field-vibration synchronization anomaly index MVo = g(Mo, Vc), where the function g is obtained through neural network training. The input is the magnetic field offset feature Mo and the vibration coupling feature Vc, and the output is the magnetic field-vibration synchronization anomaly index, which reflects the synchronization anomaly between the magnetic field and vibration.

[0084] In step S124, the vibration-temperature coupling anomaly index, the temperature-magnetic field interaction anomaly index, and the magnetic field-vibration synchronization anomaly index are respectively normalized, dynamic weight coefficients are assigned according to the contribution degrees of various anomaly indexes in historical fault data, and the normalized anomaly indexes are weighted and fused based on the dynamic weight coefficients to generate the comprehensive anomaly index set.

[0085] In the embodiments of the present disclosure, the vibration-temperature coupling anomaly index VTa, the temperature-magnetic field interaction anomaly index TMo, and the magnetic field-vibration synchronization anomaly index MVo are respectively normalized, and their values are mapped to the same value range, such as [0, 1]. Through the analysis of a large amount of historical fault data, the contribution degrees of various anomaly indexes in different fault types are determined. Assume that the contribution degree of the vibration-temperature coupling anomaly index in a certain type of fault is wVTa, the contribution degree of the temperature-magnetic field interaction anomaly index is wTMo, and the contribution degree of the magnetic field-vibration synchronization anomaly index is wMVo, and wVTa + wTMo + wMVo = 1. Dynamic weight coefficients are allocated according to these contribution degrees. The normalized anomaly indexes are weighted and fused, and the comprehensive anomaly index set I = wVTa × VTa_norm + wTMo × TMo_norm + wMVo × MVo_norm, where VTa_norm, TMo_norm, and MVo_norm are the normalized vibration-temperature coupling anomaly index, temperature-magnetic field interaction anomaly index, and magnetic field-vibration synchronization anomaly index respectively. This set comprehensively reflects the comprehensive anomaly situation of the rotor assembly.

[0086] In a possible implementation manner, in step S13, the determining the fault type set and the corresponding fault confidence score of the target rotor assembly according to the matching result between the comprehensive anomaly index set and a preset fault determination threshold includes: In step S131, the vibration-temperature coupling anomaly index is matched with a first fault threshold interval. If the matching result is beyond the first fault threshold interval, it is determined that there is a rotor imbalance fault type, and a first confidence score is calculated based on the exceeding amplitude. In the embodiments of the present disclosure, the vibration-temperature coupling anomaly index VTa is compared with a first fault threshold interval [L1, U1]. If VTa > U1 or VTa < L1, it is determined that there is a rotor imbalance fault type. A first confidence score C1 is calculated based on the exceeding amplitude. For example, if the degree of VTa exceeding the upper limit U1 is ΔVTa = VTa - U1, C1 is calculated through a preset functional relationship f1(ΔVTa). The function f1 may be a monotonically increasing function. The greater the exceeding amplitude, the higher the first confidence score C1, indicating a greater possibility of rotor imbalance fault.

[0087] In step S132, the temperature-magnetic field interaction anomaly index is matched with a second fault threshold interval. If the matching result is beyond the second fault threshold interval, it is determined that there is a bearing overheat fault type, and a second confidence score is calculated based on the exceeding amplitude. In the embodiments of the present disclosure, the temperature-magnetic field interaction anomaly index TMo is compared with the second fault threshold interval [L2, U2]. Here, L2 and U2 are boundary values determined based on a large amount of past operation data of axial magnetic field motors and bearing overheating fault cases. By analyzing the temperature-magnetic field interaction anomaly indices of many similar models of motors in normal operation and bearing overheating fault states, this reasonable interval range is obtained.

[0088] Further, when TMo is greater than U2 or less than L2, it can be determined that there is a bearing overheating fault type. Next, calculate the second confidence score C2, based on the exceeding amplitude. Assume that the exceeding amplitude is represented by ΔTMo. If TMo is greater than U2, then ΔTMo is equal to TMo minus U2; if TMo is less than L2, ΔTMo is equal to L2 minus TMo.

[0089] Further, C2 can be calculated through a specific functional relationship f2. This function f2 is determined according to the relationship between the amplitude of the temperature-magnetic field interaction anomaly index exceeding the interval in historical data and the actual occurrence probability of bearing overheating faults. For example, it may be a function obtained through multiple data fittings, and its form may be similar to a polynomial function or other function forms determined according to the characteristics of the data, aiming to make the greater the exceeding amplitude, the higher the value of C2, so as to reflect the likelihood of bearing overheating faults.

[0090] In step S133, match the magnetic field vibration synchronization anomaly index with the third fault threshold interval. If the matching result is that it exceeds the third fault threshold interval, it is determined that there is a pole shift fault type, and calculate the third confidence score based on the exceeding amplitude; In the embodiments of the present disclosure, the magnetic field vibration synchronization anomaly index MVo is compared with the third fault threshold interval [L3, U3]. L3 and U3 are also determined based on in-depth analysis of a large amount of operation monitoring data of axial magnetic field motors and records related to pole shift faults. These data cover the operation states of the motors under different working conditions and the corresponding magnetic field vibration synchronization anomaly indices.

[0091] Further, when MVo is greater than U3 or less than L3, it is determined that there is a pole shift fault type. Then calculate the third confidence score C3, and operate according to the exceeding amplitude. Let the exceeding amplitude be ΔMVo. If MVo is greater than U3, ΔMVo is equal to MVo minus U3; if MVo is less than L3, ΔMVo is equal to L3 minus MVo. Calculate C3 through the function f3. The determination method of the function f3 is similar to the previous one, and it is obtained according to the relationship between the amplitude of the magnetic field vibration synchronization anomaly index exceeding this interval in historical data and the actual occurrence probability of pole shift faults.

[0092] It can be explained that this function may have a different form from f1 and f2, but they are all designed to accurately reflect the relationship between the excess amplitude and the probability of failure, such that the greater the excess amplitude, the higher the value of C3, which indicates a greater probability of the pole offset fault occurring.

[0093] In step S134, the rotor imbalance fault type, the bearing overheating fault type, and the pole offset fault type are combined to generate the fault type set of the target rotor assembly with faults, and the first confidence score, the second confidence score, and the third confidence score are combined to generate the fault confidence score of the target rotor assembly with faults.

[0094] In the embodiments of the present disclosure, the determined rotor imbalance fault type, bearing overheating fault type, and pole offset fault type are integrated together to form a fault type set F. This set F comprehensively includes the possible fault types determined based on the matching of the previous abnormal indexes and thresholds.

[0095] Meanwhile, the first confidence score C1, the second confidence score C2, and the third confidence score C3 are combined to generate a fault confidence score C. The combination method here can be understood as a specific data combination method. For example, C1, C2, and C3 are arranged in a certain order or weighted processed in a certain way to form a comprehensive value C.

[0096] The specific weighted processing method can be determined according to the importance of different fault types to the motor during actual operation. For example, if the rotor imbalance fault has a greater impact on the running stability of the motor, then the weight of C1 may be relatively high during the determination of C; if the bearing overheating fault has a low occurrence frequency but serious consequences once it occurs, the weight of C2 can also be adjusted accordingly. In this way, the fault confidence score C can more accurately reflect the probability of the overall motor fault and the weights of various fault types therein.

[0097] It can be explained that after determining the fault type set F and the fault confidence score C, these information need to be presented in an intuitive and easy-to-understand manner, which requires generating a visual inspection report. First, design the format and layout of the report. The beginning part of the report can describe the basic information of the axial magnetic field motor rotor assembly for this inspection, such as the model of the motor, the equipment it belongs to, and the inspection time, etc.

[0098] Next, in the main body of the report, the content of the fault type set F is presented in detail. For each fault type, a concise and clear description is provided. For example, for the rotor imbalance fault type, the potential causes that may lead to this fault are described, such as uneven quality during the rotor manufacturing process or component wear caused by long-term operation. At the same time, relevant detection data is attached as evidence. Here, the data can be the vibration-temperature coupling anomaly index VTa calculated previously and its comparison with the first fault threshold interval.

[0099] For the bearing overheating fault type, the possible causes are also elaborated, such as insufficient lubrication or quality problems of the bearing itself, etc., and the matching information of the temperature-magnetic field interaction anomaly index TMo and the second fault threshold interval is shown. For the pole shift fault type, it is explained that it may be caused by improper installation or external magnetic field interference, etc., and the comparison result of the magnetic field-vibration synchronization anomaly index MVo and the third fault threshold interval is given.

[0100] After presenting the fault types, the fault confidence score C is emphasized. Explain how C is obtained by combining C1, C2, and C3, and the meaning represented by each score. For example, explain that C1 represents the confidence of the rotor imbalance fault, C2 represents the confidence of the bearing overheating fault, C3 represents the confidence of the pole shift fault, and C comprehensively reflects the possibility of the overall fault.

[0101] To make the report more intuitive, visual elements such as charts are adopted. For example, a bar chart can be drawn, with the horizontal axis representing different fault types and the vertical axis representing the corresponding confidence scores, so that it can be seen at a glance the likelihood of various fault types occurring. Or a radar chart can be used, taking different fault types as different dimensions and the fault confidence score as the radius to display the overall fault situation.

[0102] After generating the visual inspection report, it is transmitted to the target terminal for display. The target terminal can be the computer screen in the factory central control room or the mobile device held by the maintenance personnel, etc. The transmission process uses a suitable network communication protocol, such as the network transmission method based on the TCP / IP protocol. First, at the device end where the report is generated, the report data is encapsulated according to the selected protocol, and necessary header information is added, such as the IP address and port number of the target terminal. Then the encapsulated data is sent out through the network interface. On the target terminal side, the corresponding port is listened to. After receiving the data, it is unpacked according to the same protocol, the report data is extracted, and is displayed on the display interface of the terminal so that relevant personnel can obtain the fault information of the axial magnetic field motor rotor assembly in a timely manner and take corresponding maintenance measures.

[0103] An embodiment of the present disclosure also provides a fault detection system for a rotor assembly of an axial magnetic field motor. Refer to Figure 2 As shown in a construction module 210, configured to collect vibration waveform data, temperature gradient distribution data, and magnetic field intensity distribution data of the rotor assembly in the axial magnetic field motor during the operation state, and construct a dynamic monitoring data set according to the vibration waveform data, the temperature gradient distribution data, and the magnetic field intensity distribution data; a first generation module 220, configured to perform dynamic feature extraction processing on the data in the dynamic monitoring data set, generate a dynamic operation feature set including vibration coupling features, temperature abnormal fluctuation features, and magnetic field offset features, and perform multi-dimensional correlation mapping processing based on the dynamic operation feature set to generate a comprehensive abnormal index set of the rotor assembly; a determination module 230, configured to determine a fault type set and a corresponding fault confidence score of a target rotor assembly with a fault according to a matching result between the comprehensive abnormal index set and a preset fault determination threshold; a second generation module 240, configured to generate a visual detection report for the target rotor assembly according to the fault type set and the corresponding fault confidence score of the target rotor assembly, where the visual detection report is used to display a fault result on a terminal.

[0104] In a possible implementation manner, the first generation module 220 is configured to: perform frequency domain transformation processing on the vibration waveform data to generate the vibration coupling features; perform sliding window statistical analysis on the temperature gradient distribution data, calculate a temperature change trend slope corresponding to each temperature sensing unit, and use a difference value between the temperature change trend slopes of adjacent temperature sensing units as the temperature abnormal fluctuation features; perform spatial distribution fitting on the magnetic field intensity distribution data to generate an axial magnetic field intensity gradient curve, and determine the magnetic field offset features according to curvature change points of the gradient curve; perform sliding window normalization processing on the vibration coupling features, temperature abnormal fluctuation features, and magnetic field offset features respectively, and correlate the standardized features in a time-space dimension to generate the dynamic operation feature set including time sequence and spatial correlation.

[0105] In a possible implementation manner, the first generation module 220 is configured to: divide the vibration waveform data into multiple equal-length time sequence segments, and perform fast Fourier transform processing on each time sequence segment to generate corresponding vibration spectrum segments; Extract the amplitude of the fundamental frequency component, the amplitude of the harmonic component, and the high-frequency noise energy value of each vibration frequency spectrum segment; Perform a moving average process on the amplitude of the fundamental frequency component, the amplitude of the harmonic component, and the high-frequency noise energy value of multiple time series segments to generate continuous vibration frequency spectrum features; Set a first target frequency band corresponding to the resonance frequency of the rotor structure of the rotor assembly, and calculate a first proportionality coefficient of the integral value of the vibration energy in the first target frequency band to the integral value of the vibration energy in the entire frequency band; Set a second target frequency band corresponding to the bearing wear characteristic frequency of the rotor assembly, and calculate a second proportionality coefficient of the integral value of the vibration energy in the second target frequency band to the integral value of the vibration energy in the entire frequency band; Set the weight allocation ratio of the first proportionality coefficient and the second proportionality coefficient according to the historical fault data correlation between the resonance frequency of the rotor structure and the bearing wear characteristic frequency; Take the sum of the weighted first proportionality coefficient and the second proportionality coefficient as the vibration coupling characteristic.

[0106] In a possible implementation manner, the first generation module 220 is configured to: Set the length of the sliding window to a preset time period, and perform a linear regression analysis on the temperature data of each temperature sensing unit within the sliding window to obtain a temperature change trend equation; Extract the slope parameter in the temperature change trend equation as the temperature change trend slope corresponding to the temperature sensing unit; When the temperature change trend slope exceeds a preset positive change threshold, mark the temperature sensing unit as an abnormal temperature rise unit; Obtain the physical distance parameter between adjacent temperature sensing units, and calculate a temperature conduction delay compensation coefficient based on the physical distance parameter; Perform time alignment processing on the temperature change trend slopes of adjacent temperature sensing units according to the temperature conduction delay compensation coefficient; Calculate the absolute value of the slope difference between adjacent temperature sensing units after alignment, and compare the absolute value of the difference with a preset difference threshold to generate the temperature abnormal fluctuation characteristic.

[0107] In a possible implementation manner, the first generation module 220 is configured to: Construct a three-dimensional magnetic field intensity distribution surface according to the spatial coordinate positions of multiple magnetic field detection devices; Perform slicing processing on the three-dimensional distribution surface along the axial direction of the rotor assembly, and extract the magnetic field intensity distribution data of each slice plane; Perform polynomial fitting on the magnetic field intensity distribution data of each slice plane to generate a corresponding axial magnetic field intensity gradient curve; Calculate the similarity coefficient between the gradient curves of adjacent slice planes, and mark the slice planes with similarity coefficients lower than a preset threshold as magnetic field anomaly regions.

[0108] In a possible implementation, the construction module 210 is configured to Deploy a plurality of vibration sensors at preset positions of the rotor assembly, and collect the vibration waveform data through the vibration sensors with a first sampling period. The vibration waveform data includes the time series waveforms of the axial vibration component and the radial vibration component; Set a plurality of temperature sensing units on the surface of the rotor assembly, and collect the temperature gradient distribution data through the temperature sensing units with a second sampling period. The temperature gradient distribution data includes the temperature change rate of different regions and the temperature difference value between adjacent regions; Set a plurality of magnetic field detection devices along the circumferential direction of the rotor assembly, and collect the magnetic field intensity distribution data through the magnetic field detection devices with a third sampling period. The magnetic field intensity distribution data includes the axial magnetic field intensity distribution curve and the magnetic field asymmetry coefficient; Perform timestamp synchronization processing on the vibration waveform data, temperature gradient distribution data, and magnetic field intensity distribution data respectively, and use interpolation or resampling methods to unify the data with inconsistent sampling frequencies to the same time reference; Establish a data mapping table according to the spatial position relationship of the sensors, and fuse the synchronized data in a time-space alignment manner to generate the dynamic monitoring data set.

[0109] In a possible implementation, the first generation module 220 is configured to: Establish a first correlation mapping relationship between the vibration coupling characteristics and the temperature anomaly fluctuation characteristics, and calculate the vibration-temperature coupling anomaly index based on the first correlation mapping relationship; Establish a second correlation mapping relationship between the temperature anomaly fluctuation characteristics and the magnetic field offset characteristics, and calculate the temperature-magnetic field interaction anomaly index based on the second correlation mapping relationship; Establish a third correlation mapping relationship between the magnetic field offset characteristics and the vibration coupling characteristics, and calculate the magnetic field-vibration synchronization anomaly index based on the third correlation mapping relationship; Normalize the vibration-temperature coupling anomaly index, temperature-magnetic field interaction anomaly index, and magnetic field-vibration synchronization anomaly index respectively, allocate dynamic weight coefficients according to the contribution degrees of various anomaly indexes in the historical fault data, and perform weighted fusion on the normalized anomaly indexes based on the dynamic weight coefficients to generate the comprehensive anomaly index set.

[0110] In a possible implementation, the determination module 230 is configured to: Match the vibration-temperature coupling anomaly index with the first fault threshold range. If the matching result is beyond the first fault threshold range, it is determined that there is a rotor imbalance fault type, and a first confidence score is calculated based on the exceeding amplitude; Match the temperature-magnetic field interaction anomaly index with the second fault threshold range. If the matching result is beyond the second fault threshold range, it is determined that there is a bearing overheat fault type, and a second confidence score is calculated based on the exceeding amplitude; Match the magnetic field-vibration synchronization anomaly index with the third fault threshold range. If the matching result is beyond the third fault threshold range, it is determined that there is a magnetic pole offset fault type, and a third confidence score is calculated based on the exceeding amplitude; Combine the rotor imbalance fault type, the bearing overheat fault type, and the magnetic pole offset fault type to generate the fault type set of the target rotor assembly with faults, and combine the first confidence score, the second confidence score, and the third confidence score to generate the fault confidence score of the target rotor assembly with faults.

[0111] An embodiment of the present disclosure further provides an electronic device, including: A memory storing a computer program thereon; a processor for executing the computer program in the memory to implement the steps of the method according to any one of the foregoing embodiments.

[0112] Figure 3 The axial magnetic field motor rotor assembly fault detection device 100 shown includes a processor 1001 and a memory 1003. Among them, the processor 1001 and the memory 1003 are connected, such as through a bus 1002. Optionally, the axial magnetic field motor rotor assembly fault detection device 100 may further include a communication component 1004, and the communication component 1004 can be used for data interaction between the device 100 and other devices, such as data sending and / or data receiving, etc. It should be noted that in actual scheduling, the communication component 1004 is not limited to one, and the structure of the axial magnetic field motor rotor assembly fault detection device 100 does not constitute a limitation to the embodiments of the present application.

[0113] The processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logical blocks, modules, and circuits described in connection with the disclosure of this application. The processor 1001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0114] The bus 1002 may include a path for transmitting information between the above components. The bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 1002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 only a thick line is shown herein, but it does not mean that there is only one bus or one type of bus.

[0115] The memory 1003 may be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, or it may also be an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium that can be used to carry or store program code and can be read by a computer, which is not limited herein.

[0116] The memory 1003 is used to store program codes for executing the embodiments of the present disclosure, and the execution is controlled by the processor 1001. The processor 1001 is used to execute the program codes stored in the memory 1003 to implement the steps shown in the embodiment of the axial magnetic field motor rotor assembly fault detection method.

[0117] An embodiment of the present disclosure also provides a computer-readable storage medium having program code stored thereon. When the program code is executed by a processor, the steps and corresponding contents of the aforementioned embodiment of the axial magnetic field motor rotor assembly fault detection method can be implemented.

[0118] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings. However, the present disclosure is not limited to the specific details in the above embodiments. Within the technical concept of the present disclosure, various changes, modifications, replacements and variations can be made to these embodiments, and these changes, modifications, replacements and variations all fall within the scope of protection of the present disclosure.

[0119] It should also be noted that the various specific technical features described in the above specific embodiments may be combined in any suitable manner, unless there is any contradiction, and these combinations shall also be considered as the contents disclosed in this disclosure. To avoid unnecessary repetition, this disclosure will not further describe various possible combinations. The technical scope of this application is not limited to the contents of the specification and must be determined based on the scope of the claims.

Claims

1. A method for detecting faults in the rotor assembly of an axial magnetic field motor, characterized in that, The method includes: Collecting vibration waveform data, temperature gradient distribution data, and magnetic field intensity distribution data of the rotor assembly in the axial magnetic field motor during operation, and constructing a dynamic monitoring data set based on the vibration waveform data, the temperature gradient distribution data, and the magnetic field intensity distribution data; Performing dynamic feature extraction processing on the data in the dynamic monitoring data set to generate a dynamic operation feature set including vibration coupling features, temperature abnormal fluctuation features, and magnetic field offset features, and performing multi-dimensional correlation mapping processing based on the dynamic operation feature set to generate a comprehensive abnormal index set of the rotor assembly; Determining a fault type set and a corresponding fault confidence score of the target rotor assembly with a fault according to the matching result between the comprehensive abnormal index set and a preset fault determination threshold; Generating a visual inspection report for the target rotor assembly according to the fault type set and the corresponding fault confidence score of the target rotor assembly, where the visual inspection report is used to display the fault result on a terminal.

2. The method for detecting faults in the rotor assembly of an axial magnetic field motor according to claim 1, characterized in that, The performing dynamic feature extraction processing on the data in the dynamic monitoring data set to generate a dynamic operation feature set including vibration coupling features, temperature abnormal fluctuation features, and magnetic field offset features includes: Performing frequency domain transformation processing on the vibration waveform data to generate the vibration coupling features; Performing sliding window statistical analysis on the temperature gradient distribution data, calculating the temperature change trend slope corresponding to each temperature sensing unit, and using the difference value between the temperature change trend slopes of adjacent temperature sensing units as the temperature abnormal fluctuation feature; Performing spatial distribution fitting on the magnetic field intensity distribution data to generate an axial magnetic field intensity gradient curve, and determining the magnetic field offset feature according to the curvature change point of the gradient curve; Performing sliding window normalization processing on the vibration coupling features, temperature abnormal fluctuation features, and magnetic field offset features respectively, and associating the normalized features in the time-space dimension to generate the dynamic operation feature set including time series and space correlation.

3. The method for detecting the fault of the axial magnetic field motor rotor assembly according to claim 2, wherein The performing frequency domain transformation processing on the vibration waveform data to generate the vibration coupling features includes: Dividing the vibration waveform data into multiple equal-length time series segments, performing fast Fourier transform processing on each time series segment to generate a corresponding vibration spectrum segment; Extracting the fundamental frequency component amplitude, harmonic component amplitude, and high-frequency noise energy value of each vibration spectrum segment; Performing sliding average processing on the fundamental frequency component amplitude, harmonic component amplitude, and high-frequency noise energy value of multiple time series segments to generate a continuous vibration spectrum feature; Setting a first target frequency band corresponding to the resonance frequency of the rotor structure of the rotor assembly, and calculating a first proportionality coefficient of the integral value of the vibration energy in the first target frequency band to the integral value of the vibration energy in the entire frequency band; Setting a second target frequency band corresponding to the bearing wear characteristic frequency of the rotor assembly, and calculating a second proportionality coefficient of the integral value of the vibration energy in the second target frequency band to the integral value of the vibration energy in the entire frequency band; Set the weight distribution ratio of the first proportionality coefficient and the second proportionality coefficient according to the historical fault data correlation between the resonance frequency of the rotor structure and the bearing wear characteristic frequency; Take the sum of the weighted first proportionality coefficient and the second proportionality coefficient as the vibration coupling characteristic.

4. The method for detecting the fault of the rotor assembly of the axial magnetic field motor according to claim 2, wherein, The sliding window statistical analysis of the temperature gradient distribution data, calculating the temperature change trend slope corresponding to each temperature sensing unit, and taking the difference value between the temperature change trend slopes of adjacent temperature sensing units as the temperature abnormal fluctuation characteristic, includes: Set the length of the sliding window to a preset time period, and perform linear regression analysis on the temperature data of each temperature sensing unit within the sliding window to obtain a temperature change trend equation; Extract the slope parameter in the temperature change trend equation as the temperature change trend slope corresponding to this temperature sensing unit; When the temperature change trend slope exceeds the preset positive change threshold, mark this temperature sensing unit as an abnormal temperature rise unit; Obtain the physical distance parameter between adjacent temperature sensing units, and calculate the temperature conduction delay compensation coefficient based on the physical distance parameter; Perform time alignment processing on the temperature change trend slopes of adjacent temperature sensing units according to the temperature conduction delay compensation coefficient; Calculate the absolute value of the slope difference between the aligned adjacent temperature sensing units, and compare the absolute value of the difference with a preset difference threshold to generate the temperature abnormal fluctuation characteristic.

5. The method for detecting the fault of the rotor assembly of the axial magnetic field motor according to claim 2, wherein The spatial distribution fitting of the magnetic field intensity distribution data to generate an axial magnetic field intensity gradient curve, includes: Construct a three-dimensional magnetic field intensity distribution surface according to the spatial coordinate positions of multiple magnetic field detection devices; Perform slicing processing on the three-dimensional distribution surface along the axial direction of the rotor assembly, and extract the magnetic field intensity distribution data of each slice plane; Perform polynomial fitting on the magnetic field intensity distribution data of each slice plane to generate a corresponding axial magnetic field intensity gradient curve; Calculate the similarity coefficient between the gradient curves of adjacent slice planes, and mark the slice plane with a similarity coefficient lower than the preset threshold as a magnetic field abnormal area.

6. The method for detecting the fault of the axial magnetic field motor rotor assembly according to claim 1, characterized in that Collect the vibration waveform data, temperature gradient distribution data, and magnetic field intensity distribution data of the rotor assembly in the axial magnetic field motor during operation, and construct a dynamic monitoring data set according to the vibration waveform data, the temperature gradient distribution data, and the magnetic field intensity distribution data, includes: Deploy multiple vibration sensors at preset positions of the rotor assembly, and collect the vibration waveform data through the vibration sensors with a first sampling period. The vibration waveform data includes the time series waveforms of the axial vibration component and the radial vibration component; Set multiple temperature sensing units on the surface of the rotor assembly, and collect the temperature gradient distribution data through the temperature sensing units with a second sampling period. The temperature gradient distribution data includes the temperature change rates of different regions and the temperature difference values between adjacent regions; A plurality of magnetic field detection devices are arranged along the circumferential direction of the rotor assembly, and the magnetic field intensity distribution data is collected by the magnetic field detection devices with a third sampling period. The magnetic field intensity distribution data includes an axial magnetic field intensity distribution curve and a magnetic field asymmetry coefficient; Timestamp synchronization processing is respectively performed on the vibration waveform data, the temperature gradient distribution data, and the magnetic field intensity distribution data. For data with inconsistent sampling frequencies, interpolation or resampling methods are used to unify them to the same time reference; A data mapping table is established according to the spatial position relationship of the sensors, and the synchronized data is fused in a time-space alignment manner to generate the dynamic monitoring data set.

7. The method for detecting a fault of the axial magnetic field motor rotor assembly according to any one of claims 1-6, characterized in that, Performing multi-dimensional correlation mapping processing based on the dynamic operation feature set to generate a comprehensive abnormal index set of the rotor assembly, including: Establish a first correlation mapping relationship between the vibration coupling feature and the temperature abnormal fluctuation feature, and calculate a vibration-temperature coupling abnormal index based on the first correlation mapping relationship; Establish a second correlation mapping relationship between the temperature abnormal fluctuation feature and the magnetic field offset feature, and calculate a temperature-magnetic field interaction abnormal index based on the second correlation mapping relationship; Establish a third correlation mapping relationship between the magnetic field offset feature and the vibration coupling feature, and calculate a magnetic field-vibration synchronization abnormal index based on the third correlation mapping relationship; Normalize the vibration-temperature coupling abnormal index, the temperature-magnetic field interaction abnormal index, and the magnetic field-vibration synchronization abnormal index respectively. Allocate dynamic weight coefficients according to the contribution degrees of various abnormal indexes in the historical fault data, and perform weighted fusion on the normalized abnormal indexes based on the dynamic weight coefficients to generate the comprehensive abnormal index set.

8. The method for detecting the fault of the axial magnetic field motor rotor assembly according to claim 7, wherein, According to the matching result between the comprehensive abnormal index set and the preset fault determination threshold, determining the fault type set and the corresponding fault confidence score of the target rotor assembly with a fault, including: Match the vibration-temperature coupling abnormal index with the first fault threshold interval. If the matching result is beyond the first fault threshold interval, it is determined that there is a rotor imbalance fault type, and a first confidence score is calculated based on the exceeding amplitude; Match the temperature-magnetic field interaction abnormal index with the second fault threshold interval. If the matching result is beyond the second fault threshold interval, it is determined that there is a bearing overheat fault type, and a second confidence score is calculated based on the exceeding amplitude; Match the magnetic field-vibration synchronization abnormal index with the third fault threshold interval. If the matching result is beyond the third fault threshold interval, it is determined that there is a pole offset fault type, and a third confidence score is calculated based on the exceeding amplitude; Combine the rotor imbalance fault type, the bearing overheat fault type, and the pole offset fault type to generate the fault type set of the target rotor assembly with a fault, and combine the first confidence score, the second confidence score, and the third confidence score to generate the fault confidence score of the target rotor assembly with a fault.

9. A fault detection system for the rotor assembly of an axial magnetic field motor, characterized in that, The system includes: A building block configured to collect vibration waveform data, temperature gradient distribution data, and magnetic field intensity distribution data of a rotor assembly in an axial magnetic field motor during operation, and construct a dynamic monitoring data set according to the vibration waveform data, the temperature gradient distribution data, and the magnetic field intensity distribution data; A first generation module configured to perform dynamic feature extraction processing on the data in the dynamic monitoring data set, generate a dynamic operation feature set including vibration coupling features, temperature abnormal fluctuation features, and magnetic field offset features, and perform multi-dimensional correlation mapping processing based on the dynamic operation feature set to generate a comprehensive abnormal index set of the rotor assembly; A determination module configured to determine a fault type set and a corresponding fault confidence score of a target rotor assembly with a fault according to a matching result between the comprehensive abnormal index set and a preset fault determination threshold; A second generation module configured to generate a visual inspection report for the target rotor assembly according to the fault type set and the corresponding fault confidence score of the target rotor assembly, and the visual inspection report is used to display a fault result on a terminal.

10. An electronic device, characterized in that, Comprising: A memory having a computer program stored thereon; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-8.

Images