A bearing fault diagnosis method and system

Abstract

The present invention relates to the field of data processing, and the present invention relates to a bearing fault diagnosis method and system. The method includes: collecting the vibration signal of the bearing in real time during the operation of the bearing, and dividing the vibration signal into multiple segments, obtaining the current signal, acoustic signal and temperature data sequence of each vibration signal corresponding to the time period; for each vibration signal segment, calculating its corresponding noise performance degree, for each vibration signal segment, extracting its vibration characteristics, and the corresponding current signal characteristics, acoustic signal characteristics and temperature data sequence characteristics; obtaining a fused feature vector; clustering the fused feature vectors corresponding to each vibration signal segment using the FCM algorithm, and obtaining the bearing fault type based on the clustering result, and the fuzzy coefficient m is positively correlated with the noise performance degree of each vibration signal segment during clustering. The method of the present invention can diagnose the bearing fault more accurately, and the diagnostic efficiency is high.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and more specifically, to a bearing fault diagnosis method and system. Background Art

[0002] Bearings are key components in mechanical transmission systems. Their main function is to support mechanical rotating bodies, reduce friction coefficients during movement, and ensure rotation accuracy. Bearings are usually installed between the shaft and the bearing seat, allowing the shaft to rotate freely in the bearing seat and to withstand various loads (such as radial loads, axial loads, etc.) generated by the shaft during rotation. Their operating status directly affects the reliability and safety of the equipment.

[0003] The main method of bearing fault diagnosis in the prior art is to collect the vibration signal of the bearing through an acceleration sensor, extract the time domain and frequency domain characteristics of the vibration signal, and analyze the vibration signal in combination with a neural network model, a clustering algorithm or a support vector machine to determine whether the bearing has a fault and the type of fault; however, the single vibration signal analysis has limitations, such as being susceptible to noise interference and being insensitive to early weak faults, which will affect the accuracy of the diagnosis results when using neural networks or clustering algorithms to diagnose faults. For example, the rolling bearing fault diagnosis method disclosed in the Chinese patent application publication number CN119474989A only analyzes the bearing vibration signal when using a support vector machine to diagnose bearing faults, which has great limitations and leads to poor accuracy of the diagnosis results.

[0004] In order to avoid the impact of noise on the diagnostic results and improve the accuracy of the diagnostic results, the usual practice is to further identify the noise in the collected signal after collecting the signal and filter and denoise the noisy signal. However, this method will increase the amount of data processing and reduce the efficiency of bearing fault diagnosis.

[0005] In summary, the bearing fault diagnosis method in the prior art has technical problems of poor diagnostic accuracy and low diagnostic efficiency. Summary of the invention

[0006] In order to solve the technical problems of poor diagnostic accuracy and low diagnostic efficiency in the bearing fault diagnosis method in the above-mentioned prior art, the present invention provides solutions in the following aspects.

[0007] In the first aspect, the present invention provides a bearing fault diagnosis method, comprising: collecting the vibration signal of the bearing in real time during the operation of the bearing, and dividing the vibration signal into multiple segments, obtaining the current signal, acoustic signal and temperature data sequence of each vibration signal corresponding to the time period; for each vibration signal, calculating the corresponding noise performance degree according to its effective value, peak value and standard deviation, the noise performance degree of the vibration signal of the i-th segment is The calculation expression is: ; In the formula, , and Respectively represent the effective value, peak value and standard deviation of the vibration signal of the i-th segment, , and Respectively represent the Pearson correlation coefficient between the i-th segment vibration signal and the corresponding current signal, acoustic signal and temperature data sequence; For each vibration signal segment, extract its vibration features, as well as the corresponding current features of the current signal, the acoustic features of the acoustic signal, and the temperature features of the temperature data sequence; and fuse the extracted features to obtain a fused feature vector; The FCM algorithm is used to cluster the fused feature vectors corresponding to each vibration signal segment, and the fault type of the bearing is obtained based on the clustering result. During clustering, the fuzzy coefficient m is positively correlated with the noise performance of each vibration signal segment.

[0008] The beneficial effects are as follows: when the method of the present invention performs fault vibration analysis on the bearing, it no longer analyzes a single vibration signal, but analyzes signals in four dimensions: vibration, temperature, current, and acoustics. By integrating the characteristics of signals in multiple dimensions of vibration, temperature, current, and acoustics, the fault information contained in different data types is fully utilized, thereby improving the comprehensiveness and accuracy of fault diagnosis. In addition, when calculating the degree of noise expression, not only the influence of the ratio of the peak value to the effective value of the vibration signal on the degree of noise expression is considered, but also the influence of the standard deviation of the vibration signal on the degree of noise expression is considered, and the influence of the correlation between the vibration signal and the corresponding temperature signal, current signal, and acoustic signal on the degree of noise expression is also considered, so that the calculation result of the noise expression degree of the vibration signal is more accurate, further improving the accuracy of bearing fault diagnosis.

[0009] Furthermore, when using the FCM clustering algorithm to cluster the fused feature vector, the fuzzy coefficient is not set to a fixed value, but is adaptively adjusted according to the noise performance of the collected vibration signal, which improves the robustness of the algorithm to noise. Even if the noise in the collected signal is not removed, the clustering result can be guaranteed to have a high degree of accuracy, thereby improving the efficiency of bearing fault diagnosis and identification.

[0010] Preferably, the calculation expression of the fuzzy coefficient m is: ; In the formula, is the fuzzy coefficient adjustment factor, is the standard value of the fuzzy coefficient, is an exponential function with base e, Indicates the average value of the noise level of each vibration signal.

[0011] The beneficial effect is that the noise performance degree of each vibration signal is comprehensively considered when calculating the fuzzy coefficient, so that the calculation result of the fuzzy coefficient is more accurate, thereby further ensuring the robustness of the FCM clustering algorithm to noise and the accuracy of the clustering results.

[0012] Preferably, the calculation expression of the fuzzy coefficient m is: ; In the formula, represents the maximum value of the fuzzy coefficient, It indicates the average value of the noise performance of each segment of the vibration signal in the segmentation result.

[0013] Preferably, the method further comprises, after clustering, calculating the ratio of the average intra-class distance to the average inter-class distance of the clusters; In response to the ratio being greater than a preset ratio threshold, the fuzzy coefficient m is adjusted, and the adjusted fuzzy coefficient m is The calculation expression is: ; In the formula, represents the average intra-class distance, represents the average inter-class distance.

[0014] The beneficial effect is that after calculating the fuzzy coefficient based on the vibration signal, the present invention further takes into account the influence of intra-class distance and inter-class distance on clustering accuracy, and corrects the fuzzy coefficient using the ratio of average intra-class distance to average inter-class distance, thereby further improving the accuracy of bearing fault diagnosis.

[0015] Preferably, the characteristic of the current signal corresponding to the vibration signal is represented by a characteristic parameter, and its calculation expression is: ; In the formula, Represents the characteristic parameter of the current signal, represents the exponential function with base e, , and Respectively represent the current mean value, current effective value and current amplitude of the current signal, and Represent the mean and peak values ​​of the vibration signal, Indicates the harmonic content of the current signal.

[0016] Preferably, the characteristics of the acoustic signal corresponding to the vibration signal are represented by characteristic parameters, and the calculation expression thereof is: ; In the formula, Represents the characteristic parameters of the acoustic signal, represents the exponential function with base e, , and Respectively represent the mean, peak and Mel cepstral coefficients of the sound pressure level of the acoustic signal. and Represent the mean and peak values ​​of the vibration signal respectively.

[0017] Preferably, the characteristics of the temperature data sequence corresponding to the vibration signal are represented by characteristic parameters, and the calculation expression thereof is: ; In the formula, , and Respectively represent the mean, maximum value and variance of the temperature data series, Represents the characteristic parameters of the temperature data series.

[0018] Preferably, the number of clusters of the FCM algorithm is 4, and the corresponding fault states are no fault, inner ring fault, outer ring fault and rolling element fault respectively; obtaining the fault type of the bearing according to the clustering result comprises: obtaining the vibration signal of a preset number of continuous segments before the current moment from the vibration signal segmentation result as a reference vibration signal; the time length of the left end point of the time period corresponding to the reference vibration signal from the current moment is less than the preset time length; In response to each reference vibration signal falling into the same cluster, the fault state corresponding to the cluster is used as the fault diagnosis result of the bearing.

[0019] Preferably, the method for calculating the preset number of segments includes: A time period with the current moment as the right endpoint and a length equal to the preset time length is selected, and each segment of vibration signal corresponding to the time period is selected from the vibration signal segmentation result, and the preset number of segments is calculated according to the noise performance degree of each selected segment of vibration signal, and the expression is: ; In the formula, N and Respectively represent the preset number of segments and the standard value of the preset number of segments. represents the exponential function with base e, represents the peak value of the selected i-th vibration signal, n, and They respectively represent the total number of selected vibration signal segments, the mean value of noise performance, and the maximum value of noise performance. It represents the difference between the maximum peak value and the minimum peak value in each selected vibration signal.

[0020] In a second aspect, the present invention provides a bearing fault diagnosis system, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the bearing fault diagnosis method of the present invention is implemented.

[0021] In summary, the beneficial effect of the present invention is that the method of the present invention can diagnose bearing faults more accurately and with high diagnostic efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein: Figure 1 is a flow chart schematically showing a bearing fault diagnosis method according to an embodiment of the present invention; Figure 2 FIG. 1 is a schematic diagram schematically showing the structure of a bearing fault diagnosis system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0024] Bearing fault diagnosis method embodiment: like Figure 1 As shown, the bearing fault diagnosis method of the present invention comprises: S101, segmenting the collected vibration signal, and obtaining the corresponding current signal, acoustic signal and temperature data sequence, specifically: collecting the vibration signal of the bearing in real time during the operation of the bearing, and dividing the vibration signal into multiple segments, and obtaining the current signal, acoustic signal and temperature data sequence of the time period corresponding to each vibration signal segment; In this embodiment, the current signal refers to the signal of the driving current of the motor corresponding to the bearing, which can be collected by a current transformer, and the sampling frequency can be set to 1kHz to 10kHz. The vibration state of the bearing will affect the running state of the motor, thereby causing changes in the current signal. The increase in vibration intensity caused by the bearing fault will increase the load of the motor, thereby causing an increase in current. Therefore, when the bearing is subjected to fault vibration, analysis can be performed based on the current signal.

[0025] In order to obtain the current signal, acoustic signal and temperature data sequence of each vibration signal corresponding to the time period, the current signal, acoustic signal and temperature signal can be collected at the same time as the bearing vibration signal. The vibration signal of the bearing can be collected by an acceleration sensor, and the sampling frequency can be determined according to the Nyquist sampling theorem. In this embodiment, it can be set to 1kHz to 12kHz. In other embodiments, other suitable sampling frequencies can also be used. The temperature data sequence can be collected by a temperature sensor, and the sampling frequency can be set to 0.1HZ to 1HZ. The acoustic signal can be collected by a sound pressure sensor, and the sampling frequency can be set to 1kHz to 10kHz.

[0026] S102, calculating the noise performance level of each vibration signal segment, specifically: for each vibration signal segment, calculating its corresponding noise performance level according to its effective value, peak value and standard deviation, the noise performance level of the i-th vibration signal The calculation expression is: ; In the formula, , and Respectively represent the effective value, peak value and standard deviation of the vibration signal of the i-th segment, , and They represent the Pearson correlation coefficient between the i-th vibration signal and the corresponding current signal, acoustic signal and temperature data sequence respectively.

[0027] In this embodiment, the noise representation degree is used to characterize the possibility of noise existing in the vibration signal of the corresponding segment and / or the degree of influence by the noise.

[0028] Since the vibration signal acquisition frequency is originally higher than the temperature data sequence acquisition frequency, when calculating the Pearson correlation coefficient between a certain vibration signal and the corresponding temperature data sequence, the vibration acceleration at the temperature data acquisition moment in the temperature data sequence can be first screened out from the vibration signal, thereby obtaining the vibration acceleration at each temperature data acquisition moment in the temperature data sequence, and then calculating the Pearson correlation coefficient between the two.

[0029] When calculating the Pearson correlation coefficient between a certain vibration signal and the corresponding acoustic signal, we can first select multiple acceleration values ​​​​at equal intervals from the vibration signal, and then filter out the amplitude value that is the same as the selected acceleration value collection time from the acoustic signal, and then calculate the Pearson correlation coefficient between the two.

[0030] The Pearson correlation coefficient between the corresponding current signals, The peak value of a vibration signal refers to the maximum or minimum value of the vibration signal within a period of time, and the effective value of a vibration signal refers to the average level of the vibration signal within a period of time. Generally, the larger the absolute value of the ratio of the peak value to the effective value of the vibration signal, the greater the fluctuation amplitude of the vibration signal, and the greater the possibility of noise in the vibration signal. In addition, the larger the standard deviation of the vibration signal, the greater the fluctuation amplitude of the vibration signal, and the greater the possibility of noise in the vibration signal.

[0031] In addition, the vibration of the bearing during operation will increase friction, thereby generating heat and raising the temperature, that is, (the vibration signal is positively correlated with the temperature signal); in addition, the vibration caused by the bearing failure will increase the load of the motor, resulting in an increase in current, that is, (the vibration signal is positively correlated with the corresponding current signal); furthermore, the vibration caused by the bearing failure will excite sound waves of a specific frequency, increasing the high-frequency components and signal amplitude in the collected acoustic signal, that is, (the vibration signal is positively correlated with the corresponding acoustic signal). Therefore, the noise performance of the bearing can be measured based on the Pearson correlation coefficient between the vibration signal and the corresponding temperature signal, the Pearson correlation coefficient between the vibration signal and the corresponding current signal, and the Pearson correlation coefficient between the vibration signal and the corresponding acoustic signal, and the noise performance of the bearing is negatively correlated with the above three Pearson correlation coefficients.

[0032] In summary, when calculating the noise performance level of a vibration signal, this embodiment makes the noise performance level positively correlated with the corresponding absolute value of the specific gravity, positively correlated with the corresponding standard deviation, and negatively correlated with the above three Pearson correlation coefficients, so that the noise performance level of the vibration signal can be calculated more accurately.

[0033] S103, obtaining a fusion vector corresponding to each segment of the vibration signal, specifically: for each segment of the vibration signal, extracting its vibration feature, as well as the corresponding current feature of the current signal, the acoustic feature of the acoustic signal, and the temperature feature of the temperature data sequence; and fusing the extracted features to obtain a fusion feature vector; In this embodiment, the first feature of the extracted vibration signal includes a time domain feature and a frequency domain feature, wherein the time domain feature includes the effective value, peak value and standard deviation of the vibration signal, and the frequency domain feature includes the frequency mean and frequency skewness. In other embodiments, the time domain feature may further include other suitable parameters, and the frequency domain feature may further include other suitable parameters.

[0034] In this embodiment, the extracted feature of the current signal may be one or more of the current mean value, the current effective value, the current amplitude, and the harmonic content. A comprehensive feature parameter may also be calculated based on these parameters to represent the feature of the current signal.

[0035] In this embodiment, the feature of the acoustic signal extracted may be one or more of a mean value of the sound pressure level and a peak value of the sound pressure level, or a comprehensive feature parameter may be calculated based on these parameters to represent the feature of the acoustic signal.

[0036] In this embodiment, the feature of the extracted temperature data sequence may be one or more of the mean, the maximum value and the variance, or a comprehensive feature parameter may be calculated based on these parameters to represent the feature of the temperature data sequence.

[0037] When fusing features, you can use the autoencoder in deep learning to fuse the extracted multi-source features, map different types of features to the same feature space, and generate a comprehensive feature vector. You can also use concatenation, multiplication, or bilinear fusion methods to fuse the extracted features.

[0038] S104. Obtain the bearing fault type based on the fused feature vectors corresponding to each vibration signal segment. Specifically, use the FCM algorithm to cluster the fused feature vectors corresponding to each vibration signal segment, and obtain the bearing fault type based on the clustering result. During clustering, the fuzzy coefficient m is positively correlated with the noise performance of each vibration signal segment.

[0039] In the FCM algorithm, the fuzzy coefficient is usually represented by the symbol m, also known as the fuzzy index or fuzzification factor, and its value range is m>1. The fuzzy coefficient is a crucial parameter that determines the distribution of sample membership, thereby affecting the results and characteristics of clustering. The fuzzy coefficient m directly affects the distribution of membership of samples belonging to different clusters by controlling the calculation method of membership. Specifically, when the value of m is large, the distribution of membership will be flatter, that is, the probability of samples belonging to different clusters will be more uniform and the fuzziness will be stronger. On the contrary, when m is small, the distribution of membership will be sharper, and samples tend to be obviously biased towards a certain cluster, and the fuzziness is weaker. A larger fuzzy coefficient will make the membership distribution more vague, and a smaller fuzzy coefficient will make the membership distribution clearer.

[0040] Generally speaking, noise will increase the uncertainty and complexity of the vibration signal, resulting in increased ambiguity in the clustering results, and noise may make the membership distribution of the fused feature vector more dispersed, affecting the accuracy of clustering. A larger fuzzy coefficient will make the membership distribution more fuzzy, thereby reducing the impact of noise on the clustering results. Therefore, when there is more noise in the vibration signal or the probability of noise in the vibration signal is higher (that is, when the noise performance of the vibration signal is higher), appropriately increasing the fuzzy coefficient can improve the robustness of the algorithm to noise and make the clustering results more stable and accurate.

[0041] The FCM algorithm (Fuzzy C-Means algorithm) is a fuzzy clustering algorithm based on the objective function, which is mainly used for data clustering analysis.

[0042] Before using the FCM algorithm model for clustering, it is necessary to set the number of clusters and the fuzzy index of the FCM algorithm model, initialize the membership matrix of the FCM algorithm, input the fused feature vectors of each vibration signal segment into the FCM algorithm model, iteratively update the membership matrix and cluster center until the stopping criterion is met, obtain the clustering result, and then obtain the bearing fault type based on the clustering result; the fuzzy index is positively correlated with the noise performance degree of each vibration signal segment.

[0043] The number of clusters can be determined according to the number of bearing fault types, and the number of fault types is equal to the number of clusters. When the vibration signal of the bearing is first collected, the bearing is usually in normal working condition. When the collected vibration signal is clustered for the first time, the clustering sample data is small. In order to ensure that the clustering results cover the clusters corresponding to all fault types and improve the generalization ability of the model, it is necessary to collect the vibration signals of the bearings in various fault states at historical moments and the corresponding temperature data series, current signals and sound pressure signals, and perform segmented processing to ensure that there are clusters corresponding to various fault types in the clustering results. For the new sample data collected later, it can be classified into the corresponding cluster cluster based on its membership, so as to obtain its corresponding fault type.

[0044] The bearing fault diagnosis method of this embodiment, when using the FCM clustering algorithm to diagnose bearing faults, integrates the characteristics of signals in multiple dimensions such as vibration, temperature, current, and acoustics, and fully utilizes the fault information contained in different data types, thereby improving the comprehensiveness and accuracy of fault diagnosis; and can promptly detect bearing faults, provide sufficient time for equipment maintenance, and reduce equipment downtime losses and repair costs. When calculating the degree of noise performance, it is considered that the greater the absolute value of the ratio of the peak value to the effective value of the vibration signal, the higher the degree of noise performance of the vibration signal; and it is considered that the greater the standard deviation of the vibration signal, the higher the degree of noise performance of the vibration signal; and the correlation between the vibration signal and the corresponding temperature signal, current signal, and acoustic signal is taken into account, so that the calculation result of the noise performance degree of the vibration signal is more accurate, further improving the accuracy of bearing fault diagnosis; In addition, when the FCM clustering algorithm is used to cluster the fused feature vectors, the algorithm's robustness to noise is improved by adaptively adjusting the fuzzy coefficient according to the noise performance of the collected vibration signal. The accuracy of the clustering results can be guaranteed without denoising the collected vibration signal. Therefore, the method of the present invention can greatly improve the clustering efficiency while ensuring the clustering accuracy.

[0045] In one embodiment, the calculation expression of the fuzzy coefficient m is: ; In the formula, is the fuzzy coefficient adjustment factor, is the standard value of the fuzzy coefficient, is an exponential function with base e, Indicates the average value of the noise performance of each segment of the vibration signal. In this embodiment, the standard value of the fuzzy coefficient can be 1.5. The fuzzy coefficient adjustment factor can be 0.3.

[0046] In this embodiment, the noise performance level of each vibration signal is comprehensively considered when calculating the fuzzy coefficient, thereby further ensuring the robustness of the FCM clustering algorithm to noise and the accuracy of the clustering results.

[0047] In one embodiment, the calculation expression of the fuzzy coefficient m is: ; In the formula, represents the maximum value of the fuzzy coefficient, It indicates the average value of the noise performance of each segment of the vibration signal in the segmentation result. You can take 2.9.

[0048] In one embodiment, it further includes: S201, after clustering, calculating the ratio of the average intra-class distance to the average inter-class distance of the clusters; S202: In response to the ratio being greater than a preset ratio threshold, adjusting the fuzzy coefficient m. The adjusted fuzzy coefficient m is The calculation expression is: ; In the formula, represents the average intra-class distance, represents the average inter-class distance.

[0049] When the average intra-class distance is large and the average inter-class distance is small, the subsequent clustering results may overlap between different clusters, which affects the accuracy of fault diagnosis. Therefore, when the average intra-class distance is large and the average inter-class distance is small, the value of the fuzzy coefficient needs to be increased to prevent the subsequent overlap between different clusters, thereby ensuring the accuracy of fault diagnosis. In this embodiment, after the fuzzy coefficient is calculated based on the vibration signal, the fuzzy coefficient is further corrected using the ratio of the average intra-class distance to the average inter-class distance, thereby further improving the accuracy of bearing fault diagnosis.

[0050] In one embodiment, the characteristic of the current signal corresponding to the vibration signal is represented by a characteristic parameter, and its calculation expression is: ; In the formula, Represents the characteristic parameter of the current signal, represents the exponential function with base e, , and Respectively represent the current mean value, current effective value and current amplitude of the current signal, and Represent the mean and peak values ​​of the vibration signal respectively. The consistency between the fluctuation degree of the current signal and the fluctuation degree of the vibration signal can be characterized. Indicates the harmonic content of the current signal.

[0051] Different types of parameters of the current signal can characterize different aspects of characteristic information of the current signal. In the present embodiment, when calculating the characteristic parameters of the current signal, the calculation is not based on a single parameter of the current signal, but rather a comprehensive consideration of multiple types of parameters corresponding to the current signal, so that the calculated characteristic parameters can characterize various aspects of the current signal. In addition, considering that the higher the consistency between the fluctuation degree of the current signal and the fluctuation degree of the vibration signal, the less the current signal is affected by noise, the characteristic parameters of the current signal are calculated based on the consistency between the fluctuation degree of the current signal and the fluctuation degree of the vibration signal, so that the calculated characteristic parameters of the current signal can characterize the coupling relationship characteristics between the current signal and the corresponding vibration signal, thereby further improving the accuracy of the clustering results.

[0052] In one embodiment, the characteristic of the acoustic signal corresponding to the vibration signal is represented by a characteristic parameter, and its calculation expression is: ; In the formula, Represents the characteristic parameters of the acoustic signal, represents the exponential function with base e, , and Respectively represent the mean, peak and Mel cepstral coefficients of the sound pressure level of the acoustic signal. and Represent the mean and peak values ​​of the vibration signal respectively. The consistency between the fluctuation degree of the sound pressure signal and the fluctuation degree of the vibration signal can be characterized.

[0053] The mean and peak sound pressure level of the acoustic signal can reflect the overall loudness of the sound signal emitted by the bearing, and the Mel-cepstral coefficient of the acoustic signal can reflect the frequency distribution of the sound signal emitted by the bearing. The mean sound pressure level, peak and Mel-cepstral coefficient of the acoustic signal are comprehensively considered in calculating the characteristic parameters of the acoustic signal, so that the calculated characteristic parameters of the acoustic signal are more accurate. In addition, considering that the higher the consistency between the fluctuation degree of the sound pressure signal and the fluctuation degree of the vibration signal, the less the sound pressure signal is affected by noise, the characteristic parameters of the current signal are calculated based on the consistency between the fluctuation degree of the sound pressure signal and the fluctuation degree of the vibration signal, so that the calculated characteristic parameters of the sound pressure signal can characterize the coupling relationship characteristics between the sound pressure signal and the corresponding vibration signal, further improving the accuracy of the clustering results.

[0054] In one embodiment, the characteristics of the temperature data sequence corresponding to the vibration signal are represented by characteristic parameters, and the calculation expression is: ; In the formula, , and Respectively represent the mean, maximum value and variance of the temperature data series, Represents the characteristic parameters of the temperature data series. Characterizes the consistency between the fluctuation degree of the temperature data series and the fluctuation degree of the vibration signal.

[0055] When calculating the characteristic parameters of the temperature data sequence, the mean, maximum value and variance of the temperature data sequence are comprehensively considered, so that the calculated characteristic parameters of the temperature data sequence are more accurate; in addition, the characteristic parameters of the temperature data sequence are calculated based on the consistency between the fluctuation degree of the temperature data sequence and the fluctuation degree of the vibration signal, so that the calculated characteristic parameters of the temperature data sequence can characterize the coupling relationship characteristics between the temperature data sequence and the corresponding vibration signal, further improving the accuracy of the clustering results.

[0056] In one embodiment, the number of clusters of the FCM algorithm is 4, and the corresponding fault states are no fault, inner ring fault, outer ring fault and rolling element fault. The fault type of the bearing obtained according to the clustering results includes: S301, obtaining a preset number of continuous vibration signals before the current moment from the vibration signal segmentation result as a reference vibration signal; the time length from the left end point of the time period corresponding to the reference vibration signal to the current moment is less than the preset time length; For example, if the current time is 2:00 p.m., the vibration signal before the current time in the segmentation result refers to the vibration signal collected before 2:00 p.m. Assuming that the time period corresponding to the reference vibration signal is from 2:01 to 2:02, the left endpoint of the corresponding time period is 2:01.

[0057] When diagnosing a fault, the closer the vibration signal is to the current moment, the greater its reference value. Therefore, it is necessary to select a vibration signal that is closer to the current moment from the historically collected vibration signals for analysis.

[0058] S302: In response to each reference vibration signal falling into the same cluster, the fault state corresponding to the cluster is used as a fault diagnosis result of the bearing.

[0059] In the extreme case where the influence of noise on the vibration signal is too great (i.e., the sensor is subject to great interference), the corresponding fused feature vector error is large, which may cause it not to be classified into the correct clustering cluster, thereby causing the accuracy of the bearing fault diagnosis result to deteriorate; this embodiment takes into account that the noise interference on the sensor collecting the signal is usually sporadic, and when obtaining the final fault diagnosis result, the fault type is not judged based on the clustering result corresponding to a single vibration signal, but is judged based on the clustering results corresponding to multiple consecutive vibration signals, thereby weakening the influence of noise on the clustering results and improving the accuracy of fault diagnosis.

[0060] In one embodiment, the method for calculating the preset number of segments includes: A time period with the current moment as the right endpoint and a length of a preset duration is selected, and each segment of vibration signal corresponding to the time period is selected from the vibration signal segmentation result, and the preset number of segments is calculated according to the noise performance degree of each selected segment of vibration signal. The expression is: ; In the formula, N and Respectively represent the preset number of segments and the standard value of the preset number of segments. represents the exponential function with base e, represents the peak value of the selected i-th vibration signal, n, and They represent the total number of selected vibration signal segments, the mean value of noise performance, and the maximum value of noise performance, respectively. The mean value used to characterize the difference in the degree of fluctuation of two adjacent vibration signal segments. The larger the value is, the greater the frequency of change of the vibration signal in the time period is, and the more times the vibration signal is disturbed by noise, at this time the value of the preset number of segments should be as large as possible to ensure the accuracy of the fault diagnosis result; in addition, the larger the ratio of the mean value of the noise performance degree of the selected vibration signal to the maximum value of the noise performance degree is, the more times the vibration signal is disturbed by noise in the time period is, and the value of the preset number of segments should be as large as possible to ensure the accuracy of the fault diagnosis result; therefore, the preset number of segments and and The preset number of segments is calculated in direct proportion to each other, which can further improve the accuracy of the bearing fault diagnosis results.

[0061] In the selected vibration signal, the greater the difference in the fluctuation degree between two adjacent vibration signal segments, the greater the influence of noise on the vibration signal. In this embodiment, the preset duration may be 5 minutes, and in other embodiments, it may be other suitable durations.

[0062] Bearing fault diagnosis system implementation example: The present invention also provides a bearing fault diagnosis system. Figure 2 As shown, the bearing fault diagnosis system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a bearing fault diagnosis method described in the above embodiment is implemented.

[0063] The bearing fault diagnosis system also includes other components familiar to those skilled in the art, such as a communication bus and a communication interface, whose configuration and functions are known in the art and will not be described in detail here.

[0064] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0065] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A bearing fault diagnosis method, characterized in that: The method comprises: collecting the vibration signal of the bearing in real time during the operation of the bearing, dividing the vibration signal into multiple segments, obtaining the current signal, acoustic signal and temperature data sequence of each vibration signal corresponding to the time period; for each vibration signal segment, calculating the corresponding noise performance degree according to its effective value, peak value and standard deviation, and the noise performance degree of the vibration signal of the i-th segment is The calculation expression is: ; In the formula, , and Respectively represent the effective value, peak value and standard deviation of the vibration signal of the i-th segment, , and Respectively represent the Pearson correlation coefficient between the i-th segment vibration signal and the corresponding current signal, acoustic signal and temperature data sequence; For each vibration signal segment, extract its vibration features, as well as the corresponding current features of the current signal, the acoustic features of the acoustic signal, and the temperature features of the temperature data sequence; and fuse the extracted features to obtain a fused feature vector; The FCM algorithm is used to cluster the fused feature vectors corresponding to each vibration signal segment, and the fault type of the bearing is obtained based on the clustering result. During clustering, the fuzzy coefficient m is positively correlated with the noise performance of each vibration signal segment.

2. The bearing fault diagnosis method according to claim 1, characterized in that: The calculation expression of the fuzzy coefficient m is: ； In the formula, is the fuzzy coefficient adjustment factor, is the standard value of the fuzzy coefficient, is an exponential function with base e, Indicates the average value of the noise level of each vibration signal.

3. The bearing fault diagnosis method according to claim 1, characterized in that: The calculation expression of the fuzzy coefficient m is: ； In the formula, represents the maximum value of the fuzzy coefficient, It indicates the average value of the noise performance of each segment of the vibration signal in the segmentation result.

4. The bearing fault diagnosis method according to claim 2, characterized in that: It also includes, after clustering, calculating the ratio of the average intra-class distance to the average inter-class distance of the clusters; In response to the ratio being greater than a preset ratio threshold, the fuzzy coefficient m is adjusted, and the adjusted fuzzy coefficient m is The calculation expression is: ; In the formula, represents the average intra-class distance, represents the average inter-class distance.

5. The bearing fault diagnosis method according to claim 1, characterized in that: The characteristics of the current signal corresponding to the vibration signal are expressed by characteristic parameters, and the calculation expression is: ； In the formula, Represents the characteristic parameter of the current signal, represents the exponential function with base e, , and Respectively represent the current mean value, current effective value and current amplitude of the current signal, and Represent the mean and peak values ​​of the vibration signal, Indicates the harmonic content of the current signal.

6. The bearing fault diagnosis method according to claim 1, characterized in that: The characteristics of the acoustic signal corresponding to the vibration signal are expressed by characteristic parameters, and the calculation expression is: ； In the formula, Represents the characteristic parameters of the acoustic signal, represents the exponential function with base e, , and Respectively represent the mean, peak and Mel cepstral coefficients of the sound pressure level of the acoustic signal. and Represent the mean and peak values ​​of the vibration signal respectively.

7. The bearing fault diagnosis method according to claim 1, characterized in that: The characteristics of the temperature data sequence corresponding to the vibration signal are expressed by characteristic parameters, and the calculation expression is: ； In the formula, , and Respectively represent the mean, maximum value and variance of the temperature data series, Represents the characteristic parameters of the temperature data series.

8. The bearing fault diagnosis method according to any one of claims 1 to 7, characterized in that: The number of clusters of the FCM algorithm is 4, and the corresponding fault states are no fault, inner ring fault, outer ring fault and rolling element fault respectively; obtaining the fault type of the bearing according to the clustering result includes: obtaining the vibration signal of a preset number of continuous segments before the current moment from the vibration signal segmentation result as a reference vibration signal; the time length of the left end point of the time period corresponding to the reference vibration signal from the current moment is less than the preset time length; In response to each reference vibration signal falling into the same cluster, the fault state corresponding to the cluster is used as the fault diagnosis result of the bearing.

9. The bearing fault diagnosis method according to claim 8, characterized in that: The method for calculating the preset number of segments includes: A time period with the current moment as the right endpoint and a length equal to the preset time length is selected, and each segment of vibration signal corresponding to the time period is selected from the vibration signal segmentation result, and the preset number of segments is calculated according to the noise performance degree of each selected segment of vibration signal, and the expression is: ； In the formula, N and Respectively represent the preset number of segments and the standard value of the preset number of segments. represents the exponential function with base e, represents the peak value of the selected i-th vibration signal, n, and They respectively represent the total number of selected vibration signal segments, the mean value of noise performance, and the maximum value of noise performance. It represents the difference between the maximum peak value and the minimum peak value in each selected vibration signal.

10. A bearing fault diagnosis system, comprising a processor and a memory, wherein the memory stores computer program instructions, characterized in that: When the computer program instructions are executed by the processor, the bearing fault diagnosis method according to any one of claims 1 to 9 is implemented.

Images