Bearing fault diagnosis method and device, storage medium and computer equipment

Abstract

The invention provides a bearing fault diagnosis method and device, a storage medium and computer equipment. The bearing fault diagnosis method comprises the steps that an original vibration data set of a target bearing is acquired; processing the original vibration data set through each preset diagnosis model in a plurality of preset diagnosis models to determine whether each piece of original vibration data in the original vibration data set is fault data, and calculating a fault data proportion of the original vibration data set, each preset diagnosis model corresponding to a fault data proportion; obtaining the weight of each preset diagnosis model in a plurality of preset diagnosis models; and determining a fault diagnosis result of the target bearing according to the weight of each preset diagnosis model in the plurality of preset diagnosis models and the corresponding fault data proportion. According to the method, a more accurate diagnosis result can be obtained, so that the accuracy rate and the recall rate are both improved, the generalization of the method is improved, the false alarm rate and the missing report rate are reduced, and reliable diagnosis of bearing faults is guaranteed.

Description

Technical Field

[0001] The present disclosure relates to the field of wind power generation, and more specifically, to a bearing fault diagnosis method, device, storage medium and computer equipment. Background Art

[0002] In order to realize the fault diagnosis of rolling bearing vibration signals, there are many different diagnostic models in the related technologies. However, no matter which model is used, due to the different characteristics of different bearings and the differences in the application effects of the models under various working conditions, a single method has poor generalization in the actual industrial environment and high false alarm and missed alarm rates. Summary of the invention

[0003] Therefore, how to improve the generalization of fault diagnosis methods and reduce false alarm and missed alarm rates is crucial for reliable diagnosis of bearing faults.

[0004] In a general aspect, a bearing fault diagnosis method is provided, the bearing fault diagnosis method comprising: obtaining an original vibration data set of a target bearing; processing the original vibration data set through each preset diagnostic model of a plurality of preset diagnostic models to determine whether each original vibration data in the original vibration data set is fault data, and calculating the fault data ratio of the original vibration data set, wherein each preset diagnostic model corresponds to a fault data ratio; obtaining the weight of each preset diagnostic model of the plurality of preset diagnostic models; and determining the fault diagnosis result of the target bearing according to the weight of each preset diagnostic model of the plurality of preset diagnostic models and the corresponding fault data ratio.

[0005] Optionally, the number of the target bearings is at least one, wherein the processing of the original vibration data set by each preset diagnostic model among a plurality of preset diagnostic models to determine whether each original vibration data in the original vibration data set is fault data, and calculating the fault data ratio of the original vibration data set, includes: processing the original vibration data set of each of the target bearings separately by each preset diagnostic model among the plurality of preset diagnostic models to determine whether each original vibration data in the original vibration data set of each of the target bearings is fault data, and calculating the fault data ratio of the original vibration data set of each of the target bearings.

[0006] Optionally, the bearing fault diagnosis method also includes: determining the statistical value of the fault data ratio of the original vibration data set of each target bearing for each preset diagnostic model; determining the fault data ratio threshold of each preset diagnostic model according to the statistical value of the fault data ratio corresponding to each preset diagnostic model; wherein, determining the fault diagnosis result of the target bearing according to the weight of each preset diagnostic model in the multiple preset diagnostic models and the corresponding fault data ratio includes: determining the fault diagnosis result of the target bearing according to the weight of each preset diagnostic model in the multiple preset diagnostic models, the corresponding fault data ratio and the fault data ratio threshold.

[0007] Optionally, the fault diagnosis result of the target bearing is determined according to the weight of each preset diagnostic model in the multiple preset diagnostic models, the corresponding fault data ratio and the fault data ratio threshold, including: for each preset diagnostic model, determining the ratio of the fault data ratio of the target bearing to the fault data ratio threshold, and obtaining the preliminary diagnosis result of the target bearing under each preset diagnostic model; according to the weight of each preset diagnostic model in the multiple preset diagnostic models, determining the weighted average of the preliminary diagnosis results of the target bearing under the multiple preset diagnostic models, and obtaining the fault diagnosis result of the target bearing.

[0008] Optionally, determining the fault data proportion threshold of each preset diagnostic model based on the statistical value of the fault data proportion corresponding to each preset diagnostic model includes: obtaining a preset reference threshold; and taking the larger value between the preset reference threshold and the statistical value of the fault data proportion corresponding to each preset diagnostic model as the fault data proportion threshold of each preset diagnostic model.

[0009] Optionally, obtaining the weight of each preset diagnostic model among the multiple preset diagnostic models includes: obtaining a sample vibration data set, wherein the sample vibration data set includes multiple pairs of sample vibration data and fault labels, and the fault labels indicate fault or normal; processing the sample vibration data through each preset diagnostic model among the multiple preset diagnostic models to determine whether the sample vibration data is fault data as a diagnosis result; determining the accuracy and recall rate of each preset diagnostic model according to the diagnosis result and fault label of each preset diagnostic model; and calculating the sum of the accuracy and recall rate of each preset diagnostic model as the weight of each preset diagnostic model.

[0010] Optionally, the multiple preset diagnostic models include a rule diagnostic model, wherein the processing of the original vibration data set by each preset diagnostic model in the multiple preset diagnostic models to determine whether each original vibration data in the original vibration data set is fault data includes: performing time domain and frequency domain conversion processing on each original vibration data in the original vibration data set to obtain multiple pairs of vibration amplitudes and vibration frequencies corresponding to each original vibration data; extracting multiple vibration peaks from the multiple vibration amplitudes corresponding to each original vibration data by the rule diagnostic model, and recording the vibration frequencies corresponding to the multiple vibration peaks as peak frequencies; obtaining the bearing fault frequency; for each Original vibration data, when there is a peak frequency that meets a preset condition among multiple peak frequencies corresponding to the original vibration data, the original vibration data is determined to be fault data, wherein the preset condition includes: among the frequency differences formed by the peak frequency and other peak frequencies, there are at least N frequency differences, which are respectively close to at least N multiples of the bearing fault frequency, N is a preset positive integer, and the at least N multiples include 1 multiple; the preset condition also includes: the frequency difference formed by the peak frequency close to 1 multiple of the bearing fault frequency has the highest frequency of occurrence among all the frequency differences that can be formed by the multiple peak frequencies corresponding to the original vibration data.

[0011] Optionally, the at least N multiples close to the bearing fault frequency include: the difference between the at least N multiples of the bearing fault frequency is less than or equal to a set multiple of the spectral resolution, the spectral resolution is the difference between two adjacent spectral lines in the spectrum obtained after time domain and frequency domain conversion processing, and the set multiple is a positive integer.

[0012] Optionally, for each original vibration data, when there is a peak frequency that meets a preset condition among the multiple peak frequencies corresponding to the original vibration data, determining that the original vibration data is fault data includes: for each peak frequency among the multiple peak frequencies corresponding to each original vibration data, determining a first frequency difference between the peak frequency and other peak frequencies among the multiple peak frequencies corresponding to the original vibration data, and when the first frequency difference is close to 1 times the frequency of the bearing fault frequency, associating the first frequency difference with the peak frequency and saving it, to obtain multiple first frequency differences corresponding to the original vibration data, wherein each of the multiple first frequency differences is associated with at least one first frequency difference. less one peak frequency; from the multiple first frequency difference values ​​corresponding to each original vibration data, determine the first frequency difference value with the largest number of associated peak frequencies as the candidate fault frequency corresponding to each original vibration data; for each original vibration data, if there is at least one peak frequency that meets the preset sub-condition among all the peak frequencies associated with the candidate fault frequencies corresponding to the original vibration data, determine that the original vibration data is fault data, wherein the preset sub-condition includes: in the second frequency difference values ​​formed by the peak frequency and other peak frequencies among the multiple peak frequencies corresponding to the original vibration data, there are at least N second frequency difference values, which are respectively close to at least N multiples of the bearing fault frequency.

[0013] Optionally, the multiple preset diagnostic models include a shallow machine learning diagnostic model based on feature engineering, wherein the features used by the shallow machine learning diagnostic model based on feature engineering include at least one of the following: changes in the main frequency band position, and a degree of spectrum dispersion, wherein the change in the main frequency band position indicates the degree to which the main vibration frequency of the original vibration data is concentrated in high frequency or low frequency, the main vibration frequency is a vibration frequency with a relatively large vibration amplitude among multiple vibration frequencies obtained by converting the original vibration data through time domain and frequency domain, and the high frequency or low frequency is determined based on the analysis frequency of the original vibration data, and the degree of spectrum dispersion indicates the degree to which the multiple vibration frequencies obtained by converting the original vibration data through time domain and frequency domain are concentrated or dispersed relative to the main vibration frequency.

[0014] Optionally, the multiple preset diagnostic models include deep learning models.

[0015] In another general aspect, a bearing fault diagnosis device is provided, comprising: an acquisition unit configured to acquire an original vibration data set of a target bearing; a diagnosis unit configured to process the original vibration data set through each preset diagnostic model in a plurality of preset diagnostic models to determine whether each original vibration data in the original vibration data set is fault data, and calculate the fault data ratio of the original vibration data set, wherein each preset diagnostic model corresponds to a fault data ratio; the acquisition unit is further configured to acquire a weight of each preset diagnostic model in the plurality of preset diagnostic models; and a fusion unit configured to determine a fault diagnosis result of the target bearing according to the weight of each preset diagnostic model in the plurality of preset diagnostic models and the corresponding fault data ratio.

[0016] Optionally, the number of the target bearings is at least one, wherein the diagnostic unit is further configured to process the original vibration data set of each target bearing separately through each preset diagnostic model among the multiple preset diagnostic models to determine whether each original vibration data in the original vibration data set of each target bearing is fault data, and calculate the fault data ratio of the original vibration data set of each target bearing.

[0017] Optionally, the bearing fault diagnosis device also includes a statistical unit, which is configured to: determine the statistical value of the fault data ratio of the original vibration data set of each target bearing for each preset diagnostic model; determine the fault data ratio threshold of each preset diagnostic model according to the statistical value of the fault data ratio corresponding to each preset diagnostic model; wherein the fusion unit is further configured to determine the fault diagnosis result of the target bearing according to the weight of each preset diagnostic model in the multiple preset diagnostic models, the corresponding fault data ratio and the fault data ratio threshold.

[0018] Optionally, the fusion unit is further configured to: determine, for each preset diagnostic model, the ratio of the fault data proportion of the target bearing to the fault data proportion threshold, and obtain the preliminary diagnostic result of the target bearing under each preset diagnostic model; and determine, according to the weight of each preset diagnostic model in the multiple preset diagnostic models, the weighted average of the preliminary diagnostic results of the target bearing under the multiple preset diagnostic models, and obtain the fault diagnosis result of the target bearing.

[0019] Optionally, the statistical unit is further configured to: obtain a preset reference threshold; and use the larger value of the preset reference threshold and the statistical value of the fault data proportion corresponding to each preset diagnostic model as the fault data proportion threshold of each preset diagnostic model.

[0020] Optionally, the acquisition unit is further configured to: acquire a sample vibration data set, wherein the sample vibration data set includes multiple pairs of sample vibration data and fault labels, and the fault labels indicate fault or normal; process the sample vibration data through each preset diagnostic model of the multiple preset diagnostic models to determine whether the sample vibration data is fault data as a diagnostic result; determine the accuracy and recall rate of each preset diagnostic model according to the diagnostic result and fault label of each preset diagnostic model; calculate the sum of the accuracy and recall rate of each preset diagnostic model as the weight of each preset diagnostic model.

[0021] Optionally, the multiple preset diagnostic models include a rule diagnostic model, wherein the diagnostic unit is further configured to: perform time domain and frequency domain conversion processing on each original vibration data in the original vibration data set to obtain multiple pairs of vibration amplitudes and vibration frequencies corresponding to each original vibration data; extract multiple vibration peaks from the multiple vibration amplitudes corresponding to each original vibration data through the rule diagnostic model, and record the vibration frequencies corresponding to the multiple vibration peaks as peak frequencies; obtain the bearing fault frequency; for each original vibration data, if there is a peak frequency that meets a preset condition among the multiple peak frequencies corresponding to the original vibration data, determine that the original vibration data is fault data, wherein the preset condition includes: among the frequency differences formed by the peak frequency and other peak frequencies, there are at least N frequency differences, which are respectively close to at least N multiples of the bearing fault frequency, N is a preset positive integer, and the at least N multiples include 1 multiple; the preset condition also includes: the frequency difference formed by the peak frequency that is close to 1 multiple of the bearing fault frequency has the highest frequency of occurrence among all the frequency differences that can be formed by the multiple peak frequencies corresponding to the original vibration data.

[0022] Optionally, the at least N multiples close to the bearing fault frequency include: the difference between the at least N multiples of the bearing fault frequency is less than or equal to a set multiple of the spectral resolution, the spectral resolution is the difference between two adjacent spectral lines in the spectrum obtained after time domain and frequency domain conversion processing, and the set multiple is a positive integer.

[0023] Optionally, the diagnostic unit is further configured to: for each peak frequency among multiple peak frequencies corresponding to each original vibration data, determine a first frequency difference between the peak frequency and other peak frequencies among the multiple peak frequencies corresponding to the original vibration data, and when the first frequency difference is close to 1 times the frequency of the bearing fault frequency, associate the first frequency difference with the peak frequency and save it to obtain multiple first frequency differences corresponding to the original vibration data, wherein each of the multiple first frequency differences is associated with at least one peak frequency; from the multiple first frequency differences corresponding to each original vibration data, determine a first frequency difference with the largest number of associated peak frequencies as a candidate fault frequency corresponding to each original vibration data; for each original vibration data, when there is at least one peak frequency that satisfies a preset sub-condition among all the peak frequencies associated with the candidate fault frequency corresponding to the original vibration data, determine that the original vibration data is fault data, wherein the preset sub-condition includes: among the second frequency differences formed by the peak frequency and other peak frequencies among the multiple peak frequencies corresponding to the original vibration data, there are at least N second frequency differences, which are respectively close to at least N times the frequency of the bearing fault frequency.

[0024] Optionally, the multiple preset diagnostic models include a shallow machine learning diagnostic model based on feature engineering, wherein the features used by the shallow machine learning diagnostic model based on feature engineering include at least one of the following: changes in the main frequency band position, and a degree of spectrum dispersion, wherein the change in the main frequency band position indicates the degree to which the main vibration frequency of the original vibration data is concentrated in high frequency or low frequency, the main vibration frequency is a vibration frequency with a relatively large vibration amplitude among multiple vibration frequencies obtained by converting the original vibration data through time domain and frequency domain, and the high frequency or low frequency is determined based on the analysis frequency of the original vibration data, and the degree of spectrum dispersion indicates the degree to which the multiple vibration frequencies obtained by converting the original vibration data through time domain and frequency domain are concentrated or dispersed relative to the main vibration frequency.

[0025] Optionally, the multiple preset diagnostic models include deep learning models.

[0026] In another general aspect, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by at least one processor, the at least one processor is prompted to perform the bearing fault diagnosis method as described above.

[0027] In another general aspect, a computer device is provided, comprising: at least one processor; and at least one memory storing computer executable instructions, wherein the computer executable instructions, when executed by the at least one processor, cause the at least one processor to execute the bearing fault diagnosis method as described above.

[0028] The present disclosure proposes a bearing fault diagnosis method, device, storage medium and computer equipment, which can effectively reflect the difference in diagnosis results of different preset diagnostic models by calculating the proportion of fault data obtained after diagnosis by different preset diagnostic models. On this basis, by combining the model weights to fuse these fault data proportions, the fusion results are used to reflect the comprehensive fault diagnosis results, which can weaken the difference in the effects of different models when facing different bearings and different working conditions, and obtain more accurate diagnosis results compared to a single model, thereby improving both precision and recall, which helps to improve the generalization of the method, reduce false alarm rates and missed alarm rates, and ensure reliable diagnosis of bearing faults.

[0029] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The above and other objects and features of the present invention will become more apparent through the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0031] Figure 1 is a flow chart showing a bearing fault diagnosis method according to an embodiment of the present disclosure;

[0032] Figure 2 is a flow chart showing a bearing fault diagnosis method according to a specific embodiment of the present disclosure;

[0033] Figure 3 is a schematic diagram showing a diagnosis process of a rule diagnosis model according to a specific embodiment of the present disclosure;

[0034] Figure 4 is a partial structural schematic diagram of a wind turbine generator set showing a diagnosis example according to the present disclosure;

[0035] Figure 5 is a block diagram showing a bearing fault diagnosis device according to an embodiment of the present disclosure;

[0036] Figure 6 is a block diagram illustrating a computer device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0037] The following specific embodiments are provided to help the reader obtain a comprehensive understanding of the methods, devices and / or systems described herein. However, after understanding the disclosure of the present application, various changes, modifications and equivalents of the methods, devices and / or systems described herein will be clear. For example, the order of operations described herein is only an example and is not limited to those orders set forth herein, but can be changed as will be clear after understanding the disclosure of the present application, except for operations that must occur in a specific order. In addition, for greater clarity and simplicity, the description of features known in the art may be omitted.

[0038] The features described herein can be implemented in different forms and should not be construed as being limited to the examples described herein. Rather, the examples described herein have been provided to illustrate only some of the many possible ways to implement the methods, devices, and / or systems described herein, which will be clear after understanding the disclosure of the present application.

[0039] As used herein, the term "and / or" includes any one of the associated listed items and any combination of any two or more.

[0040] Although terms such as "first", "second", and "third" may be used herein to describe various members, components, regions, layers, or portions, these members, components, regions, layers, or portions should not be limited by these terms. Instead, these terms are only used to distinguish one member, component, region, layer, or portion from another member, component, region, layer, or portion. Therefore, without departing from the teachings of the examples described herein, the first member, first component, first region, first layer, or first portion referred to in the examples may also be referred to as the second member, second component, second region, second layer, or second portion.

[0041] In the specification, when an element (such as a layer, a region, or a substrate) is described as being “on”, “connected to”, or “coupled to” another element, the element may be directly “on”, “connected to”, or “coupled to” another element, or one or more other elements may be present therebetween. Conversely, when an element is described as being “directly on”, “directly connected to”, or “directly coupled to” another element, there may be no other elements present therebetween.

[0042] The terms used herein are only used to describe various examples and are not intended to limit the disclosure. Unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. The terms "comprise", "include" and "have" indicate the presence of the described features, quantities, operations, components, elements and / or combinations thereof, but do not exclude the presence or addition of one or more other features, quantities, operations, components, elements and / or combinations thereof.

[0043] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by a person of ordinary skill in the art to which the present disclosure belongs after understanding the present disclosure. Unless explicitly defined as such herein, terms (such as those defined in a general dictionary) should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and the present disclosure, and should not be interpreted in an idealized or overly formal manner.

[0044] Furthermore, in the description of examples, when it is considered that a detailed description of a well-known related structure or function would cause vague interpretation of the present disclosure, such a detailed description will be omitted.

[0045] Figure 1 is a flow chart illustrating a bearing fault diagnosis method according to an embodiment of the present disclosure.

[0046] Reference Figure 1 , in step S101, an original vibration data set of a target bearing is obtained.

[0047] In step S102, the original vibration data set is processed by each preset diagnostic model in a plurality of preset diagnostic models to determine whether each original vibration data in the original vibration data set is fault data, and the fault data ratio of the original vibration data set is calculated, wherein each preset diagnostic model corresponds to a fault data ratio. This step uses a plurality of preset diagnostic models to perform preliminary diagnosis on the original vibration data set respectively, and obtains the corresponding fault data ratio as the preliminary diagnosis result of the corresponding preset diagnostic model, which is reserved for use in subsequent steps. It should be understood that the larger the fault data ratio, the higher the possibility that the target bearing has a fault based on the corresponding preset diagnostic model.

[0048] In step S103, the weight of each preset diagnostic model in the plurality of preset diagnostic models is obtained. The weight can reflect the reference value of each preset diagnostic model.

[0049] Optionally, step S103 includes: obtaining a sample vibration data set, wherein the sample vibration data set includes multiple pairs of sample vibration data and fault labels, and the fault labels indicate fault or normal; processing the sample vibration data through each preset diagnostic model in multiple preset diagnostic models to determine whether the sample vibration data is fault data as a diagnostic result; determining the accuracy and recall rate of each preset diagnostic model according to the diagnostic result and fault label of each preset diagnostic model; calculating the sum of the accuracy and recall rate of each preset diagnostic model as the weight of each preset diagnostic model. The accuracy and recall rate can effectively reflect the diagnostic effect of the preset diagnostic model from different angles. By using the sum of the two as the weight of the corresponding model, the weight can represent the diagnostic effect of the model, and then determine the reference value of the preliminary diagnostic result of the corresponding model in the fusion process according to the diagnostic effect, which helps to improve the accuracy and recall of the fused fault diagnosis results and ensure the reliable diagnosis of bearing faults.

[0050] In step S104, the fault diagnosis result of the target bearing is determined according to the weight of each preset diagnostic model in the multiple preset diagnostic models and the corresponding fault data ratio. This step fuses the initial diagnostic results of the multiple preset diagnostic models based on the weights, for example, the weighted average of the fault data ratios corresponding to the multiple preset diagnostic models can be calculated as the final fault diagnosis result.

[0051] According to the bearing fault diagnosis method of the exemplary embodiment of the present disclosure, by calculating the proportion of fault data obtained after diagnosis by different preset diagnostic models, the difference in diagnostic results of different preset diagnostic models can be effectively reflected with the help of one data. On this basis, by combining the model weights to fuse these fault data proportions, the fusion results are used to reflect the comprehensive fault diagnosis results, which can weaken the difference in the effects of different models when facing different bearings and different working conditions, and obtain more accurate diagnosis results compared to a single model, thereby improving both precision and recall, which helps to improve the generalization of the method, reduce false alarm rates and missed alarm rates, and ensure reliable diagnosis of bearing faults.

[0052] Next, we will introduce in detail how to integrate the preliminary diagnostic results of multiple preset diagnostic models.

[0053] Optionally, the number of target bearings is at least one, wherein step S102 includes: processing the original vibration data set of each target bearing respectively through each preset diagnostic model in a plurality of preset diagnostic models to determine whether each original vibration data in the original vibration data set of each target bearing is fault data, and calculating the fault data ratio of the original vibration data set of each target bearing. By designing the number of target bearings to be at least one and performing initial diagnosis on each target bearing respectively, batch diagnosis of multiple target bearings can be achieved, which helps to improve diagnostic efficiency.

[0054] Further optionally, the bearing fault diagnosis method according to the embodiment of the present disclosure also includes: for each preset diagnostic model, determining the statistical value (such as mean value, mode, specific quantile, etc.) of the fault data ratio of the original vibration data set of each target bearing; determining the fault data ratio threshold of each preset diagnostic model according to the statistical value of the fault data ratio corresponding to each preset diagnostic model; wherein step S104 includes: determining the fault diagnosis result of the target bearing according to the weight of each preset diagnostic model in multiple preset diagnostic models, the corresponding fault data ratio and the fault data ratio threshold. Although theoretically, the fault data ratio can reflect the possibility of fault in the target bearing, strictly speaking, for different batches of model diagnosis calculations, even if the fault data ratio of two target bearings is the same, the actual probability of fault between the two is still different, and there is a diagnostic difference caused by the target bearing itself. By statistically calculating the statistical value of the fault data ratio of all target bearings obtained by a preset diagnostic model in this diagnosis, it can be used as a reference value for this diagnosis, thereby reducing the error caused by the difference between different bearings in each diagnosis, and improving the reliability of the diagnosis result. It should be understood that for a preset diagnostic model, if the proportion of fault data of a target bearing under the model exceeds the corresponding fault data proportion threshold, it can be considered that the target bearing is diagnosed to have a fault under the model. As an example, the diagnosed target bearing may include multiple bearings in multiple units. At this time, in order to simplify data processing, bearing fault diagnosis can also be performed for each unit, that is, data processing is no longer performed on a single target bearing as a unit, but on a single unit as a unit, and the data of multiple target bearings in the unit are regarded as the data of the same target bearing, and the proportion of fault data therein is calculated, and diagnostic processing is performed. This is also an implementation method of the present disclosure.

[0055] It should also be understood that the more target bearings / units are diagnosed in the current batch, the better the reference effect of the statistical value of the fault data ratio is; when the number of target bearings / units is small, the reference effect of the statistical value is reduced. In particular, when the number of target bearings / units is 1, the fault data ratio is only 1, and the statistical value of the fault data ratio is most likely the fault data ratio itself. Therefore, for the case where the reference effect of these statistical values ​​is insufficient, additional data processing can be performed to improve the reference effect of the fault data ratio threshold.

[0056] As an example, a sufficiently large preset reference threshold can be configured. Accordingly, the above-mentioned step of determining the threshold of the proportion of fault data of each preset diagnostic model according to the statistical value of the proportion of fault data corresponding to each preset diagnostic model includes: obtaining the preset reference threshold; using the larger value of the preset reference threshold and the statistical value of the proportion of fault data corresponding to each preset diagnostic model as the threshold of the proportion of fault data of each preset diagnostic model. By adding a sufficiently large preset reference threshold and using the larger value of this value and the statistical value as the threshold of the proportion of fault data, it can be ensured that the threshold of the proportion of fault data used is always greater than or equal to the preset reference threshold, thereby providing a lower threshold limit, which can compensate for the situation of insufficient statistical data, ensure the reference effect of the threshold of the proportion of fault data obtained, and reduce the risk of the target bearing being misdiagnosed as a fault. It should be understood that the preset reference threshold can be obtained through experience or through experiments, and the present disclosure does not limit this.

[0057] Optionally, the above-mentioned step of determining the fault diagnosis result of the target bearing according to the weight of each preset diagnostic model in multiple preset diagnostic models, the corresponding fault data ratio and the fault data ratio threshold includes: for each preset diagnostic model, determining the ratio of the fault data ratio of the target bearing to the fault data ratio threshold, and obtaining the preliminary diagnosis result of the target bearing under each preset diagnostic model; according to the weight of each preset diagnostic model in multiple preset diagnostic models, determining the weighted average of the preliminary diagnosis results of the target bearing under multiple preset diagnostic models, and obtaining the fault diagnosis result of the target bearing. Specifically, how to determine the fault diagnosis result of the target bearing in combination with the fault data ratio threshold of each preset diagnostic model, by first calculating the ratio of the fault data ratio to the fault data ratio threshold of the corresponding preset diagnostic model as the preliminary diagnosis result of the model, is equivalent to using the fault data ratio threshold to perform dimensionless processing on the fault data ratio, and when the preliminary diagnosis result is greater than or equal to 1, it can be considered that the target bearing is diagnosed to have a fault under the model. At this time, after calculating the weighted average of the preliminary diagnostic results of multiple preset diagnostic models, the final fault diagnosis result of the target bearing can still be determined based on this principle according to the size relationship between the weighted average and 1, without the need for other processing, thereby achieving convenient fault diagnosis using simple data processing.

[0058] Next, different preset diagnostic models are introduced respectively.

[0059] In some embodiments, optionally, the plurality of preset diagnostic models include a rule diagnostic model. Accordingly, the step of processing the original vibration data set by each preset diagnostic model in the plurality of preset diagnostic models in step S102 to determine whether each original vibration data in the original vibration data set is fault data includes: performing time domain and frequency domain conversion processing on each original vibration data in the original vibration data set to obtain multiple pairs of vibration amplitudes and vibration frequencies corresponding to each original vibration data; extracting multiple vibration peaks from the multiple vibration amplitudes corresponding to each original vibration data by the rule diagnostic model, and recording the vibration frequencies corresponding to the multiple vibration peaks as peak frequencies; obtaining Take the bearing fault frequency; for each original vibration data, if there is a peak frequency that meets the preset conditions among the multiple peak frequencies corresponding to the original vibration data, determine that the original vibration data is fault data, wherein the preset conditions include: in the frequency difference formed by the peak frequency and other peak frequencies, there are at least N frequency differences, which are respectively close to at least N times the bearing fault frequency, N is a preset positive integer, and at least N times include 1 times. The preset conditions also include: the frequency difference formed by the peak frequency close to 1 times the bearing fault frequency has the highest frequency among all the frequency differences that can be formed by the multiple peak frequencies corresponding to the original vibration data. By adopting a rule-based diagnosis model based on peak extraction and frequency recognition, the physical characteristics at the time of the fault can be used to construct the preset conditions satisfied by the fault data, thereby realizing fault diagnosis, reducing the amount of data calculation, and ensuring that the diagnosis results have a physical theoretical basis. Subsequently, optimized diagnosis can be achieved by optimizing the preset conditions. Specifically, the bearing fault frequency is calculated based on the insulation diameter, tilt angle, and number of rollers of the bearing, and is used to reflect the theoretical frequency when a specific part of the bearing fails. In other words, for faults in different parts, there are often different bearing fault frequencies, so the relationship between the vibration data of the bearing and the frequency multiples of the bearing fault frequency corresponding to a certain part can be detected to diagnose whether the part has a fault. However, since the actual operation of the bearing is still affected by many factors, there is a discrepancy between the actual frequency when a fault occurs and the theoretically calculated bearing fault frequency. The frequency difference between the two peak frequencies can reflect the vibration situation. If the frequency difference of a peak frequency is close to multiple multiples of the bearing fault frequency (including 1 multiple), and the frequency difference formed by the peak frequency is close to the bearing fault frequency many times, then the frequency difference formed by this peak frequency close to the bearing fault frequency is considered to be the actual fault frequency, and the bearing is determined to have a fault accordingly. As an example, N = 3. It should be noted that since frequency domain data (i.e., frequency spectrum, including multiple pairs of vibration amplitudes and vibration frequencies) are usually used in vibration analysis, and the data directly detected by the sensor is often time domain data, it is necessary to first perform time-domain to frequency domain conversion on the detected time domain data to obtain frequency domain data, and then input the frequency domain data into the model for diagnosis.The conversion processing steps here can be regarded as data preparation before formally using each preset diagnostic model for preliminary diagnosis.

[0060] Further optionally, close to at least N multiples of the bearing fault frequency, including: the difference with at least N multiples of the bearing fault frequency is less than or equal to the set multiple of the spectrum resolution, the spectrum resolution is the difference between two adjacent spectrum lines in the spectrum obtained after the time domain and frequency domain conversion process, and the set multiple is a positive integer. By clarifying "close" as the difference is less than or equal to the set multiple of the spectrum resolution, a clear basis for judgment can be provided, which is helpful to realize the automatic diagnosis of the computer. In addition, the spectrum resolution reflects the accuracy of the time domain and frequency domain conversion process. By specifically using the set multiple of the spectrum resolution as a reference for judgment, the judgment can be adapted to the actual time domain and frequency domain conversion process, thereby providing a more reasonable basis for judgment when facing different time domain and frequency domain conversion processes. It should be understood that the set multiple can be obtained in combination with experiments or based on experience, and the present disclosure is not limited to this. As an example, the multiple is set to 2.

[0061] Further optionally, for each original vibration data, when there is a peak frequency that meets a preset condition among the multiple peak frequencies corresponding to the original vibration data, determining that the original vibration data is fault data includes: for each peak frequency among the multiple peak frequencies corresponding to each original vibration data, determining a first frequency difference between the peak frequency and other peak frequencies among the multiple peak frequencies corresponding to the original vibration data, and when the first frequency difference is close to 1 times the frequency of the bearing fault frequency, associating the first frequency difference with the peak frequency and saving it, to obtain multiple first frequency differences corresponding to the original vibration data, wherein each of the multiple first frequency differences is associated with at least one Peak frequency; from the multiple first frequency differences corresponding to each original vibration data, determine the first frequency difference with the largest number of associated peak frequencies as the candidate fault frequency corresponding to each original vibration data; for each original vibration data, if there is at least one peak frequency that satisfies the preset sub-condition among all the peak frequencies associated with the candidate fault frequency corresponding to the original vibration data, determine that the original vibration data is fault data, wherein the preset sub-condition includes: in the second frequency difference formed by the peak frequency and other peak frequencies among the multiple peak frequencies corresponding to the original vibration data, there are at least N second frequency differences, which are respectively close to at least N times the frequency of the bearing fault frequency. The first two steps can ensure that the candidate fault frequency obtained thereby has the highest frequency of occurrence among all the frequency differences that can be formed by the multiple peak frequencies corresponding to the original vibration data, which is equivalent to satisfying the second condition in the preset conditions. At the same time, since the candidate fault frequency is close to 1 times the frequency of the bearing fault frequency, each peak frequency associated with the candidate fault frequency can satisfy the 1 times frequency condition in the first condition of the preset conditions. On this basis, the third step can be directly executed for all peak frequencies associated with the candidate fault frequency to determine whether the candidate fault frequency meets the first condition in the preset conditions, and it is no longer necessary to determine whether each peak frequency meets the 1-time frequency condition. This judgment order can effectively reduce the amount of calculation and improve the diagnostic efficiency while meeting the judgment needs.

[0062] As an example, specifically, after peak extraction, a peak list can be obtained, which includes multiple peak frequencies. First, the peak list is traversed, and for the current peak frequency A i , calculate the frequencies B of all other peaks in the peak list j (j≠i) and A i The difference between j -A i - bearing fault frequency)≤2 times the spectrum resolution, it can be considered that (B j -A i ) is close to the bearing failure frequency, and the (B j -A i ) and Ai Save it. After traversing, all (BA)s and their corresponding A that meet the condition can be obtained. Since (BA) is a difference, there may be multiple A's corresponding to the same (BA). The (BA) with the most corresponding A's is taken as the candidate fault frequency. Then, a similar method is used to traverse the multiple A's corresponding to the candidate fault frequency. The difference is that only the bearing fault frequency (i.e. 1 times the frequency) is used to determine (B j -A i ) is close to 1 times the frequency, here multiple times that are less than the highest frequency are judged separately. For example, the highest frequency obtained after the time domain and frequency domain conversion is between 4 times and 5 times the frequency, then here the frequency is between 1 times and 4 times. If at least one of the multiple A's traversed can meet the judgment condition of more than 3 times the frequency, for example, there are 3 (BA) of a certain A that are close to 1 times, 2 times, and 4 times the frequency respectively, then the candidate fault frequency is considered to be the actual fault frequency, and the target bearing fault is determined.

[0063] In some embodiments, optionally, multiple preset diagnostic models include a shallow machine learning diagnostic model based on feature engineering, wherein the features used by the shallow machine learning diagnostic model based on feature engineering include at least one of the following: main frequency band position change, spectrum concentration and dispersion, wherein the main frequency band position change indicates the degree to which the main vibration frequency of the original vibration data is concentrated in high frequency or low frequency, the main vibration frequency is the vibration frequency with a relatively large vibration amplitude among the multiple vibration frequencies obtained by the original vibration data through time domain and frequency domain conversion processing, the high frequency or low frequency is determined based on the analysis frequency of the original vibration data, and the spectrum concentration and dispersion degree indicates the degree to which the multiple vibration frequencies obtained by the original vibration data through time domain and frequency domain conversion processing are concentrated or dispersed relative to the main vibration frequency. By adopting a shallow machine learning diagnostic model based on feature engineering and introducing at least one of the main frequency band position change and spectrum concentration and dispersion degree as the features used by the model, the diagnostic effect of the model can be effectively improved. Specifically, the main frequency band position change reflects the component characteristics of the main vibration frequency, and the spectrum concentration and dispersion degree reflects the distribution relationship between other vibration frequencies and the main vibration frequency, and the two can reflect the frequency characteristics of the original vibration data from different aspects. It should be understood that for the change in the position of the main frequency band, high frequency and low frequency refer to relatively high and relatively low frequencies within the analysis frequency range of the vibration data. The minimum analysis frequency is 0, and the maximum is the value obtained by dividing the sampling frequency by 2.56 (2 in textbooks, and 2.56 is used in the industry). For example, when the sampling frequency is 1280Hz, the maximum analysis frequency is 500Hz, so the analysis frequency is related to the sampling frequency. The data exceeding the maximum value is greatly affected by other factors and has low accuracy, so it is not analyzed. Accordingly, low-pass filtering will be performed during the time domain and frequency domain conversion processing to filter out components with extremely high frequencies.

[0064] As an example, the main frequency band position change is calculated by the following formula:

[0065]

[0066] Among them, s k is the frequency amplitude sequence of the original vibration data, k = 1, 2, 3, ..., K is the number of spectrum lines; f k is the frequency value of the kth spectral line. The larger the value of fe1, the more concentrated the main vibration frequency is at high frequencies, and vice versa.

[0067] The degree of spectrum concentration is calculated by the following formula:

[0068]

[0069] The meaning of each symbol in the formula can be found above. The smaller the value of fe2, the more vibration frequencies there are around the main vibration frequency in the data, and the more concentrated the spectrum is. Otherwise, the more dispersed it is.

[0070] It should be understood that in addition to the changes in the main frequency band position and the degree of spectrum dispersion, the shallow machine learning diagnostic model based on feature engineering can also use other conventional features to utilize sufficient information to ensure reliable diagnosis.

[0071] Optionally, the multiple preset diagnostic models include a deep learning model. The deep learning model has a deep structure and strong nonlinear feature extraction capability, and can directly extract fault features and identify fault patterns in the original vibration data, which is conducive to end-to-end fault diagnosis under complex working conditions. By using a deep learning model, the selection range of the preset diagnostic model can be broadened, which helps to improve the final diagnostic effect.

[0072] Next, combine Figures 2 to 5 , introduces a bearing fault diagnosis method according to a specific embodiment of the present disclosure.

[0073] In this specific embodiment, the preset diagnostic models used include a mechanism-based rule diagnostic model, a feature engineering-based XGBoost diagnostic model, and a deep learning-based CNN (Convolutional Neural Network) diagnostic model. First, the XGBoost diagnostic model and the CNN diagnostic model are pre-trained in advance based on the historical fault case library. There are multiple historical label data in the historical fault case library, that is, historical vibration data with labels. The labels indicate whether the bearing corresponding to the data is faulty or normal. Each data can be directly stored in the form of a spectrum to reduce the amount of data storage, and the spectrum can be directly used when pre-training the model. The trained model is used as the preset diagnostic model. These preset diagnostic models do not update parameters or train during diagnosis, but only make judgments on the data. Fault diagnosis is performed on multiple target bearings. The data in the original vibration data set uses transmission chain vibration monitoring data, and the detected time domain data can be uniformly converted to the time domain and frequency domain. For example, the FFT (fast Fourier transform) method is used for processing to obtain frequency domain data. Each preset diagnostic model uses this frequency domain data as input. See the process. Figure 2 .

[0074] Specifically, the XGBoost diagnostic model and the CNN diagnostic model are first pre-trained. The specific training method is a mature technology in this field and will not be elaborated here. After the pre-training is completed, the sum of the precision and recall of each preset diagnostic model is calculated as the weight of the model.

[0075] Then, before the formal diagnosis, the original vibration data to be diagnosed can be tested for data quality, that is, data cleaning. The difference between the data cleaning here and the conventional data cleaning is that for the special form of vibration, the abnormal data that does not repeatedly change around 0 needs to be eliminated. Data cleaning mainly involves the calculation of some characteristic indicators, including mean deviation (used to detect whether the data is offset near 0), the ratio of the number of data points with the same value to the total data volume (not too much), the ratio of the difference between the positive and negative data points to the total data points (reflecting whether the data changes evenly around 0), positive impact of vibration data (indicating data that is too large and exceeds the normal vibration limit, for example, the amplitude range is 0.1-0.5. If the amplitude of a certain data is above 10, it needs to be eliminated), drift anomaly threshold (zero crossing rate, similar to positive impact of vibration data), vibration data interruption (indicating that the size of the data suddenly changes cliff-like), sliding window mean difference (if the difference between the mean of the values ​​in the first sliding window and the mean of the values ​​in the last sliding window is large, it indicates that the data has undergone a large change and does not conform to the characteristics of vibration data. For example, the mean of the first 500 data is -0.5, and the mean of the last 500 data is 0.5, and the difference is -1, which is considered abnormal). It also includes the commonly used features of vibration data, such as effective values, kurtosis values, and the proportion of null values ​​and non-real values ​​in vibration data. Only data that passes the data quality test will enter the diagnosis stage. Otherwise, it will be put into the database as part of the original vibration data and participate in the subsequent result statistics.

[0076] Next, a preliminary diagnosis is made using various preset diagnostic models.

[0077] Reference to the diagnostic process of the rule diagnosis model Figure 3 , using a diagnostic method based on peak extraction and frequency recognition. In simple terms, the diagnostic logic is to extract the peak value of the spectrum, and search for the multiple frequency that satisfies the difference with the bearing fault frequency less than 2 times the spectrum resolution in the peak difference formed by the corresponding peak list. When the number of multiple frequency is greater than or equal to 3, it is considered to be a fault, and specifically the fault is the part corresponding to the searched bearing fault frequency. Figure 3 As shown, peak extraction specifically extracts two types of data, one is the large value that ranks in the top 0.2% of all vibration frequencies in the spectrum, and the other is the value that is greater than the sum of the mean and variance, where the mean and variance refer to the mean and variance of all vibration frequencies in the spectrum.

[0078] Regarding the XGBoost diagnostic model, before inputting data into the trained XGBoost diagnostic model, the data spectrum must be processed first. If it is too long, it must be downsampled, and if it is insufficient, it must be padded with zeros to meet the input of a spectrum with a length of 8192 and a maximum analysis frequency of 500Hz. Several eigenvalues ​​are calculated and screened out for the processed spectrum, and then the eigenvalues ​​are input into the XGBoost diagnostic model. The model result output is 0 or 1, 0 indicates normal, and 1 indicates fault.

[0079] Regarding the CNN diagnostic model, before the data is input into the trained CNN diagnostic model, the data spectrum also needs to be processed as mentioned above, and then the processed spectrum is input into the CNN diagnostic model, and the model result output is also 0 or 1.

[0080] Subsequently, the output values ​​of each model are counted, and the statistical model can be used. In the preliminary diagnosis process, the statistical model will record the number of data with qualified quality by unit, calculate the proportion of fault data of each unit diagnosed by different preset diagnostic models, and use max(0.36, (avg+3*std)) as the threshold of fault data proportion. Among them, avg is the mean of the proportion of fault data of all units, std is the standard deviation of the proportion of fault data of all units, (avg+3*std) represents the statistical value of the proportion of fault data, and 0.36 is the preset reference threshold.

[0081] Finally, the diagnostic results of each preset diagnostic model are summarized.

[0082] Assume that for a target bearing, the proportion of fault data of model i is prop i , the fault data ratio threshold is thre i , with weight w i , the final model result is:

[0083]

[0084] When result is greater than or equal to 1, the target bearing is judged to be faulty.

[0085] In a diagnostic example, based on the bearing fault diagnosis method of the above-mentioned specific embodiment, the present disclosure diagnoses the bearings of a batch of units. The original vibration data set contains vibration acceleration data of 744 GW2X units in 51 wind farms from 2022 to 2023, including 736 positive cases (normal units) and 8 negative cases (problem units). IEPE (Integrated Electronics Piezo-Electric) acceleration sensors are installed near the main bearing and stator. The acceleration sensor at the main bearing position is a low-frequency acceleration sensor with a frequency range of 0.1Hz-10000Hz. The acceleration sensor at the stator position is a wideband acceleration sensor with a frequency range of 0.5Hz-10000Hz. There are 7 acceleration sensors in total, which are respectively set as follows: Figure 4 The front bearing 41 shown (the front bearing 41 is located at Figure 4, and thus its position is indicated by a dotted lead line), at the rear bearing 43 and the stator 44. Specifically, the front bearing 41 is provided with two sensors for detecting radial acceleration, which are respectively arranged at a horizontal position (3 o'clock or 9 o'clock) and a vertical position (12 o'clock or 6 o'clock) to detect the radial acceleration in the horizontal direction and the radial acceleration in the vertical direction; the rear bearing 43 is also provided with a sensor for detecting radial acceleration in the horizontal position and a sensor for detecting radial acceleration in the vertical position; the stator 44 is provided with two sensors for detecting radial acceleration and a sensor for detecting axial acceleration, the former are respectively arranged at two horizontal positions, namely, one at the 3 o'clock position and one at the 9 o'clock position, and the latter is ensured to be able to detect axial acceleration, and can be arranged at the edge, for example, also at the 3 o'clock position. The acceleration data used in this example are mainly the acceleration data of the main bearing measuring point. The data contains vibration acceleration waveforms of two acquisition strategies. Acquisition definition 1 has 65536 sampling points, a sampling frequency of 25600 Hz, and a sampling period of 2.56 s. Acquisition definition 2 has 65536 sampling points, a sampling frequency of 1280 Hz, and a sampling period of 51.2 s.

[0086] The features selected by feature engineering are as follows:

[0087] freq_spectrum_80_180_c3: group entropy of 80Hz-180Hz;

[0088] freq_spectrum_20_120_index_mass_quantile: quantile index of 20Hz-120Hz;

[0089] freq_spectrum_number_cwt_peaks_I: number of peaks;

[0090] freq_spectrum_longest_strike_above_mean: the longest continuous self-sequence length above the mean;

[0091] freq_spectrum_energy_ratio_by_chunks_8: chunk local entropy ratio;

[0092] freq_spectrum_340_440_var: variance of 340Hz-440Hz;

[0093] freq_spectrum_8: main frequency band position change;

[0094] freq_spectrum_0_100_energy_ratio_by_chunks_0: chunked local entropy ratio of 0Hz-100Hz;

[0095] freq_spectrum_12: degree of spectrum concentration;

[0096] freq_spectrum_0_100_agg_linear_trend: 0Hz-100Hz linear regression based on block time series aggregation values;

[0097] freq_spectrum_first_location_of_minimum: minimum value location;

[0098] freq_spectrum_longest_strike_below_mean: The longest continuous sequence length below the mean.

[0099] The test results are shown in the table below.

[0100] Model False Positives False negatives Real failure Accuracy Recall Rule Diagnosis Model 10 0 8 44.44％ 100.00％ CNN Diagnosis Model 10 2 8 37.50％ 75.00％ XGBoost Diagnostic Model 13 1 8 35.00％ 87.50％ Fusion Model 3 0 8 72.73％ 100.00％

[0101] In the actual wind turbine operation and maintenance scenario, a wind turbine main shaft bearing fault diagnosis model is built to provide timely warnings before major components fail, accurately analyze the cause of the failure after the failure, and formulate reasonable maintenance plans and spare parts allocation plans, which can greatly reduce operation and maintenance costs and increase power generation. At present, single models are mainly used in actual application scenarios. Physical models, statistical models, and knowledge models each have their own advantages, but single types are difficult to avoid the limitations of models. For example, in the test results, the rule-based diagnosis model of 744 wind turbines misreported 10 units, the CNN-based diagnosis model misreported 10 units and missed 2 units, and the XGBoost-based diagnosis model misreported 13 units and missed 1 unit. This disclosure proposes a fusion solution for multi-model results, which weights the results of multiple preset diagnosis models. For example, in the test results, the fusion model is a weighted fusion of the results of the rule diagnosis model, the CNN diagnosis model, and the XGBoost diagnosis model, which ultimately reduces the number of false alarms to 3 units and the number of missed alarms to 0 units. Compared with the three preset diagnosis models, the fusion model has significantly improved precision and recall. The above examples prove that the fusion solution can improve the model diagnosis effect, thereby more accurately diagnosing the main shaft bearing status of wind turbines and providing more effective support for planned operation and maintenance.

[0102] Figure 5 is a block diagram showing a bearing fault diagnosis device according to an embodiment of the present disclosure.

[0103] Reference Figure 5The bearing fault diagnosis device 500 includes an acquisition unit 501 , a diagnosis unit 502 , and a fusion unit 503 .

[0104] The acquisition unit 501 may acquire an original vibration data set of a target bearing.

[0105] The diagnostic unit 502 can process the original vibration data set through each preset diagnostic model among multiple preset diagnostic models to determine whether each original vibration data in the original vibration data set is fault data, and calculate the fault data ratio of the original vibration data set, wherein each preset diagnostic model corresponds to a fault data ratio.

[0106] The acquisition unit 501 may also acquire the weight of each preset diagnosis model in the plurality of preset diagnosis models.

[0107] Optionally, the acquisition unit 501 can also: acquire a sample vibration data set, wherein the sample vibration data set includes multiple pairs of sample vibration data and fault labels, and the fault labels indicate fault or normal; process the sample vibration data through each preset diagnostic model of multiple preset diagnostic models to determine whether the sample vibration data is fault data as a diagnosis result; determine the accuracy and recall rate of each preset diagnostic model according to the diagnostic result and fault label of each preset diagnostic model; calculate the sum of the accuracy and recall rate of each preset diagnostic model as the weight of each preset diagnostic model.

[0108] The fusion unit 503 may determine the fault diagnosis result of the target bearing according to the weight of each preset diagnosis model in the plurality of preset diagnosis models and the corresponding fault data proportion.

[0109] Optionally, the number of target bearings is at least one, wherein the diagnostic unit 502 may also process the original vibration data set of each target bearing separately through each preset diagnostic model in a plurality of preset diagnostic models to determine whether each original vibration data in the original vibration data set of each target bearing is fault data, and calculate the fault data ratio of the original vibration data set of each target bearing.

[0110] Optionally, the bearing fault diagnosis device also includes a statistical unit (not shown in the figure), which can: determine the statistical value of the fault data ratio of the original vibration data set of each target bearing for each preset diagnostic model; determine the fault data ratio threshold of each preset diagnostic model according to the statistical value of the fault data ratio corresponding to each preset diagnostic model; wherein the fusion unit 503 can also determine the fault diagnosis result of the target bearing according to the weight of each preset diagnostic model in multiple preset diagnostic models, the corresponding fault data ratio and the fault data ratio threshold.

[0111] Optionally, the fusion unit 503 may also: determine, for each preset diagnostic model, the ratio of the fault data proportion of the target bearing to the fault data proportion threshold, and obtain the preliminary diagnostic result of the target bearing under each preset diagnostic model; determine the weighted average of the preliminary diagnostic results of the target bearing under multiple preset diagnostic models according to the weight of each preset diagnostic model in multiple preset diagnostic models, and obtain the fault diagnosis result of the target bearing.

[0112] Optionally, the statistical unit may also: obtain a preset reference threshold; and use the larger value of the preset reference threshold and the statistical value of the fault data proportion corresponding to each preset diagnostic model as the fault data proportion threshold of each preset diagnostic model.

[0113] Optionally, multiple preset diagnostic models include a rule diagnostic model, wherein the diagnostic unit 502 can also: perform time domain and frequency domain conversion processing on each original vibration data in the original vibration data set through the rule diagnostic model to obtain multiple pairs of vibration amplitudes and vibration frequencies corresponding to each original vibration data; extract multiple vibration peaks from the multiple vibration amplitudes corresponding to each original vibration data, and record the vibration frequencies corresponding to the multiple vibration peaks as peak frequencies; obtain the bearing fault frequency; for each original vibration data, if there is a peak frequency that meets the preset conditions among the multiple peak frequencies corresponding to the original vibration data, determine that the original vibration data is fault data, wherein the preset conditions include: in the frequency difference formed by the peak frequency and other peak frequencies, there are at least N frequency differences, which are respectively close to at least N multiples of the bearing fault frequency, N is a preset positive integer, and at least N multiples include 1 multiple; the preset conditions also include: the frequency difference formed by the peak frequency close to 1 multiple of the bearing fault frequency has the highest frequency among all the frequency differences that can be formed by the multiple peak frequencies corresponding to the original vibration data.

[0114] Optionally, it is close to at least N multiples of the bearing fault frequency, including: the difference between at least N multiples of the bearing fault frequency is less than or equal to a set multiple of the spectral resolution, the spectral resolution is the difference between two adjacent spectral lines in the spectrum obtained after time domain and frequency domain conversion processing, and the set multiple is a positive integer.

[0115] Optionally, the diagnostic unit 502 may also: determine, for each peak frequency among the multiple peak frequencies corresponding to each original vibration data, a first frequency difference between the peak frequency and other peak frequencies among the multiple peak frequencies corresponding to the original vibration data, and when the first frequency difference is close to 1 times the frequency of the bearing fault frequency, associate the first frequency difference with the peak frequency and save it to obtain multiple first frequency differences corresponding to the original vibration data, wherein each of the multiple first frequency differences is associated with at least one peak frequency; determine, from the multiple first frequency differences corresponding to each original vibration data, a first frequency difference with the largest number of associated peak frequencies as a candidate fault frequency corresponding to each original vibration data; for each original vibration data, when there is at least one peak frequency that satisfies a preset sub-condition among all the peak frequencies associated with the candidate fault frequencies corresponding to the original vibration data, determine that the original vibration data is fault data, wherein the preset sub-condition includes: among the second frequency differences formed by the peak frequency and other peak frequencies among the multiple peak frequencies corresponding to the original vibration data, there are at least N second frequency differences, which are respectively close to at least N times the frequency of the bearing fault frequency.

[0116] Optionally, multiple preset diagnostic models include a shallow machine learning diagnostic model based on feature engineering, wherein the features used by the shallow machine learning diagnostic model based on feature engineering include at least one of the following: changes in the main frequency band position, and the degree of spectrum dispersion, wherein the change in the main frequency band position indicates the degree to which the main vibration frequency of the original vibration data is concentrated in the high frequency or low frequency, the main vibration frequency is a vibration frequency with a relatively large vibration amplitude among multiple vibration frequencies obtained by converting the original vibration data through time domain and frequency domain, and the high frequency or low frequency is determined based on the analysis frequency of the original vibration data, and the degree of spectrum dispersion indicates the degree to which the multiple vibration frequencies obtained by converting the original vibration data through time domain and frequency domain are concentrated or dispersed relative to the main vibration frequency.

[0117] Optionally, the multiple preset diagnostic models include a deep learning model.

[0118] Regarding the device in the above embodiment, the specific manner in which each unit performs the operation has been described in detail in the embodiment of the method, and will not be elaborated here.

[0119] The bearing fault diagnosis method according to the embodiment of the present disclosure can be written as a computer program and stored on a computer-readable storage medium. When the instructions corresponding to the computer program are executed by the processor, the bearing fault diagnosis method as described above can be implemented. Examples of computer-readable storage media include: read-only memory (ROM), random access programmable read-only memory (PROM), electrically erasable programmable read-only memory (EEPROM), random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), flash memory, non-volatile memory, CD-ROM, CD-R, CD+R, CD-RW, CD+RW, DVD-ROM, DVD-R, DVD+R, DVD-RW, DVD+RW, DVD-RAM, BD-ROM, BD-R, BD-R, B-R, B-RW ... D-RE, Blu-ray or optical disk storage, hard disk drive (HDD), solid state drive (SSD), card storage (such as, multimedia card, secure digital (SD) card or extreme digital (XD) card), magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid state disk and any other device, any other device is configured to store computer program and any associated data, data files and data structures in a non-transitory manner and provide the computer program and any associated data, data files and data structures to a processor or computer so that the processor or computer can execute the computer program. In one example, the computer program and any associated data, data files and data structures are distributed on a networked computer system so that the computer program and any associated data, data files and data structures are stored, accessed and executed in a distributed manner by one or more processors or computers.

[0120] Figure 6 is a block diagram illustrating a computer device according to an embodiment of the present disclosure.

[0121] Reference Figure 6 The computer device 600 includes at least one memory 601 and at least one processor 602, wherein the at least one memory 601 stores a set of computer executable instructions. When the set of computer executable instructions is executed by the at least one processor 602, a bearing fault diagnosis method according to an exemplary embodiment of the present disclosure is executed.

[0122] As an example, the computer device 600 may be a PC, a tablet device, a personal digital assistant, a smart phone, or other device capable of executing the above instruction set. Here, the computer device 600 is not necessarily a single electronic device, but may also be any device or circuit collection capable of executing the above instructions (or instruction sets) individually or in combination. The computer device 600 may also be part of an integrated control system or system manager, or may be configured as a portable electronic device interconnected with a local or remote (e.g., via wireless transmission) interface.

[0123] In computer device 600, processor 602 may include a central processing unit (CPU), a graphics processing unit (GPU), a programmable logic device, a dedicated processor system, a microcontroller, or a microprocessor. By way of example and not limitation, the processor may also include an analog processor, a digital processor, a microprocessor, a multi-core processor, a processor array, a network processor, etc.

[0124] The processor 602 may execute instructions or codes stored in the memory 601, which may also store data. Instructions and data may also be sent and received over a network via a network interface device, which may employ any known transmission protocol.

[0125] The memory 601 may be integrated with the processor 602, for example, by placing RAM or flash memory within an integrated circuit microprocessor or the like. In addition, the memory 601 may include a separate device, such as an external disk drive, a storage array, or any other storage device that can be used by a database system. The memory 601 and the processor 602 may be operatively coupled, or may communicate with each other, such as through an I / O port, a network connection, etc., so that the processor 602 can read files stored in the memory.

[0126] In addition, the computer device 600 may also include a video display (such as a liquid crystal display) and a user interaction interface (such as a keyboard, a mouse, a touch input device, etc.) All components of the computer device 600 may be connected to each other via a bus and / or a network.

[0127] The present disclosure provides a bearing fault diagnosis method, device, storage medium and computer equipment, which can effectively reflect the difference in diagnosis results of different preset diagnostic models by calculating the proportion of fault data obtained after diagnosis by different preset diagnostic models. On this basis, by combining the model weights to fuse these fault data proportions, the fusion results are used to reflect the comprehensive fault diagnosis results, which can weaken the difference in the effects of different models when facing different bearings and different working conditions, and obtain more accurate diagnosis results compared to a single model, thereby improving both the precision and recall rates, helping to improve the generalization of the method, reduce the false alarm rate and missed alarm rate, and ensure reliable diagnosis of bearing faults.

[0128] The specific implementation methods of the present disclosure have been described in detail above. Although some embodiments have been shown and described, those skilled in the art should understand that these embodiments may be modified and varied without departing from the principles and spirit of the present disclosure, the scope of which is defined by the claims and their equivalents. These modifications and variations should also be within the scope of protection of the claims of the present disclosure.

Claims

1. A bearing fault diagnosis method, characterized in that: The bearing fault diagnosis method comprises: Obtaining a set of raw vibration data of a target bearing; Processing the original vibration data set by each preset diagnostic model of a plurality of preset diagnostic models to determine whether each original vibration data in the original vibration data set is fault data, and calculating the fault data ratio of the original vibration data set, wherein each preset diagnostic model corresponds to a fault data ratio; Obtaining a weight of each preset diagnostic model in the plurality of preset diagnostic models; The fault diagnosis result of the target bearing is determined according to the weight of each preset diagnostic model in the multiple preset diagnostic models and the corresponding fault data ratio.

2. The bearing fault diagnosis method according to claim 1, characterized in that: The number of the target bearing is at least one, wherein the processing of the original vibration data set by each preset diagnostic model in a plurality of preset diagnostic models to determine whether each original vibration data in the original vibration data set is fault data, and calculating the fault data ratio of the original vibration data set includes: The original vibration data set of each target bearing is processed separately by each preset diagnostic model among the multiple preset diagnostic models to determine whether each original vibration data in the original vibration data set of each target bearing is fault data, and the fault data ratio of the original vibration data set of each target bearing is calculated.

3. The bearing fault diagnosis method according to claim 2, characterized in that: The bearing fault diagnosis method further comprises: For each preset diagnostic model, determining a statistical value of a proportion of fault data in the original vibration data set of each target bearing; Determine a threshold value of the fault data ratio for each preset diagnostic model according to the statistical value of the fault data ratio corresponding to each preset diagnostic model; Wherein, determining the fault diagnosis result of the target bearing according to the weight of each preset diagnostic model in the plurality of preset diagnostic models and the corresponding fault data proportion includes: The fault diagnosis result of the target bearing is determined according to the weight of each preset diagnostic model in the multiple preset diagnostic models, the corresponding fault data proportion and the fault data proportion threshold.

4. The bearing fault diagnosis method according to claim 3, characterized in that: The determining of the fault diagnosis result of the target bearing according to the weight of each preset diagnostic model in the plurality of preset diagnostic models, the corresponding fault data proportion and the fault data proportion threshold value comprises: For each preset diagnostic model, determine the ratio of the fault data proportion of the target bearing to the fault data proportion threshold, and obtain a preliminary diagnostic result of the target bearing under each preset diagnostic model; According to the weight of each preset diagnostic model in the multiple preset diagnostic models, a weighted average value of the preliminary diagnostic results of the target bearing under the multiple preset diagnostic models is determined to obtain a fault diagnosis result of the target bearing.

5. The bearing fault diagnosis method according to claim 3, characterized in that: The determining of the fault data proportion threshold of each preset diagnostic model according to the statistical value of the fault data proportion corresponding to each preset diagnostic model includes: Obtaining a preset reference threshold; The larger value between the preset reference threshold and the statistical value of the fault data proportion corresponding to each preset diagnostic model is used as the fault data proportion threshold of each preset diagnostic model.

6. The bearing fault diagnosis method according to claim 1, characterized in that: The obtaining of the weight of each preset diagnostic model in the plurality of preset diagnostic models comprises: Acquire a sample vibration data set, wherein the sample vibration data set includes a plurality of pairs of sample vibration data and fault labels, and the fault labels indicate fault or normal; Processing the sample vibration data by each of the plurality of preset diagnostic models to determine whether the sample vibration data is fault data as a diagnosis result; Determine the accuracy and recall rate of each preset diagnostic model based on the diagnostic results and fault labels of each preset diagnostic model; The sum of the precision and recall of each preset diagnostic model is calculated as the weight of each preset diagnostic model.

7. The bearing fault diagnosis method according to claim 1, characterized in that: The plurality of preset diagnostic models include a rule diagnostic model, wherein the processing of the original vibration data set by each preset diagnostic model in the plurality of preset diagnostic models to determine whether each original vibration data in the original vibration data set is fault data comprises: Performing time-domain and frequency-domain conversion processing on each original vibration data in the original vibration data set to obtain multiple pairs of vibration amplitudes and vibration frequencies corresponding to each original vibration data; Extracting multiple vibration peaks from multiple vibration amplitudes corresponding to each original vibration data using the rule diagnosis model, and recording the vibration frequencies corresponding to the multiple vibration peaks as peak frequencies; Get bearing failure frequency; For each original vibration data, if there is a peak frequency that meets a preset condition among the multiple peak frequencies corresponding to the original vibration data, the original vibration data is determined to be fault data, wherein the preset condition includes: among the frequency differences formed by the peak frequency and other peak frequencies, there are at least N frequency differences, which are respectively close to at least N multiples of the bearing fault frequency, N is a preset positive integer, and the at least N multiples include 1 multiple; the preset condition also includes: the frequency difference formed by the peak frequency that is close to 1 multiple of the bearing fault frequency has the highest frequency of occurrence among all the frequency differences that can be formed by the multiple peak frequencies corresponding to the original vibration data.

8. The bearing fault diagnosis method according to claim 7, characterized in that: The at least N multiples close to the bearing fault frequency include: the difference between the at least N multiples of the bearing fault frequency is less than or equal to a set multiple of the spectral resolution, the spectral resolution is the difference between two adjacent spectral lines in the spectrum obtained after time domain and frequency domain conversion processing, and the set multiple is a positive integer.

9. The bearing fault diagnosis method according to claim 7, characterized in that: For each piece of original vibration data, when there is a peak frequency that satisfies a preset condition among a plurality of peak frequencies corresponding to the original vibration data, determining that the original vibration data is fault data includes: For each peak frequency among the multiple peak frequencies corresponding to each original vibration data, determine a first frequency difference between the peak frequency and other peak frequencies among the multiple peak frequencies corresponding to the original vibration data, and when the first frequency difference is close to 1 times the frequency of the bearing fault frequency, associate the first frequency difference with the peak frequency and save it, so as to obtain a plurality of first frequency differences corresponding to the original vibration data, wherein each of the plurality of first frequency differences is associated with at least one peak frequency; Determine, from among the multiple first frequency difference values ​​corresponding to each original vibration data, a first frequency difference value having the largest number of associated peak frequencies as a candidate fault frequency corresponding to each original vibration data; For each original vibration data, if there is at least one peak frequency that satisfies preset sub-conditions among all peak frequencies associated with the candidate fault frequency corresponding to the original vibration data, the original vibration data is determined to be fault data, wherein the preset sub-conditions include: among the second frequency differences formed by the peak frequency and other peak frequencies among the multiple peak frequencies corresponding to the original vibration data, there are at least N second frequency differences, which are respectively close to at least N multiples of the bearing fault frequency.

10. The bearing fault diagnosis method according to claim 1, characterized in that: The multiple preset diagnostic models include a shallow machine learning diagnostic model based on feature engineering, wherein the features used by the shallow machine learning diagnostic model based on feature engineering include at least one of the following: changes in the main frequency band position and the degree of spectrum dispersion, wherein the change in the main frequency band position indicates the degree to which the main vibration frequency of the original vibration data is concentrated in the high frequency or low frequency, the main vibration frequency is a vibration frequency with a relatively large vibration amplitude among multiple vibration frequencies obtained by converting the original vibration data through time domain and frequency domain, and the high frequency or low frequency is determined based on the analysis frequency of the original vibration data, and the degree of spectrum dispersion indicates the degree to which the multiple vibration frequencies obtained by converting the original vibration data through time domain and frequency domain are concentrated or dispersed relative to the main vibration frequency.

11. The bearing fault diagnosis method according to claim 1, characterized in that: The multiple preset diagnostic models include a deep learning model.

12. A bearing fault diagnosis device, characterized in that: The bearing fault diagnosis device comprises: An acquisition unit, configured to acquire a raw vibration data set of a target bearing; A diagnosis unit is configured to process the original vibration data set through each preset diagnosis model of a plurality of preset diagnosis models to determine whether each original vibration data in the original vibration data set is fault data, and calculate a fault data ratio of the original vibration data set, wherein each preset diagnosis model corresponds to a fault data ratio; The acquisition unit is further configured to acquire a weight of each preset diagnostic model in the plurality of preset diagnostic models; The fusion unit is configured to determine the fault diagnosis result of the target bearing according to the weight of each preset diagnostic model in the multiple preset diagnostic models and the corresponding fault data ratio.

13. A computer device, characterized in that: include: at least one processor; at least one memory storing computer executable instructions, Wherein, when the computer executable instructions are executed by the at least one processor, the at least one processor is prompted to execute the bearing fault diagnosis method according to any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that: When the instructions in the computer-readable storage medium are executed by at least one processor, the at least one processor is prompted to perform the bearing fault diagnosis method according to any one of claims 1 to 11.

Images