A micro-precision bearing fatigue damage detection system and a detection method

Abstract

The application relates to the technical field of bearing fatigue damage detection, and discloses a micro-precision bearing fatigue damage detection system and a detection method, which comprises the following steps: collecting ultrasonic wave data of a micro-precision bearing, decomposing different high-frequency sub-band components, calculating initial threshold values and noise degrees of the high-frequency sub-band components, calculating information similarity and information retention degrees of the high-frequency sub-band components, using a wavelet threshold denoising algorithm to denoise the ultrasonic wave data according to the initial threshold values, the noise degrees and the information retention degrees of each high-frequency sub-band component, and obtaining a fatigue damage detection result according to the denoised ultrasonic wave data. The application can improve the precision of micro-precision bearing fatigue damage detection.

Description

Technical Field

[0001] This invention relates to the field of bearing fatigue damage detection technology, specifically to a micro-precision bearing fatigue damage detection system and method. Background Technology

[0002] Miniature precision bearings are commonly used in critical fields such as aerospace, medical devices, and precision instruments. Fatigue damage can lead to sudden bearing failure and equipment malfunction. Ultrasonic waves can be used to detect fatigue damage in miniature precision bearings, allowing for the timely identification and intervention of potential problems before an accident occurs. However, ultrasonic waves have short propagation paths and rapid attenuation in tiny components, often resulting in significant absorption and scattering losses in the material. Furthermore, this reduces the signal-to-noise ratio of the acquired ultrasonic data, introducing a large amount of noise.

[0003] Wavelet thresholding denoising algorithms can be used to denoise ultrasonic data. However, wavelet thresholding denoising algorithms generally select fixed constant values ​​as thresholds based on experience, which can easily remove effective information as noise during denoising, affecting the accuracy of fatigue damage detection in micro-precision bearings. Summary of the Invention

[0004] This invention provides a fatigue damage detection system and method for micro-precision bearings to solve the problem of inaccurate fatigue damage detection results due to poor noise reduction of ultrasonic data from micro-precision bearings. The specific technical solution adopted is as follows: In a first aspect, one embodiment of the present invention provides a method for detecting fatigue damage in micro-precision bearings, the method comprising the following steps: Ultrasonic data of micro-precision bearings are collected, and the ultrasonic data is decomposed into low-frequency sub-band components and different high-frequency sub-band components. The initial threshold of the high-frequency sub-band components is calculated. Based on the differences in all the data contained in the high-frequency sub-band component, the interference degree of the high-frequency sub-band component is determined. Based on the differences in the upper and lower envelope curves of the high-frequency sub-band component before and after smoothing, the non-smoothness of the high-frequency sub-band component is calculated. Based on the interference degree and non-smoothness of the high-frequency sub-band component, the noise content of the high-frequency sub-band component is determined. Any high-frequency sub-band component is denoted as the target high-frequency sub-band component. Based on the difference between the reconstruction results of the ultrasonic data of all high-frequency sub-band components other than the target high-frequency sub-band component and the acquired ultrasonic data, the information similarity of the target high-frequency sub-band component is calculated. The information similarity of all high-frequency sub-band components is clustered. Based on the number of clusters obtained by clustering and the difference in information similarity within different clusters, the information retention degree of each high-frequency sub-band component is obtained. The wavelet threshold denoising algorithm is used to denoise the ultrasonic data based on the initial threshold, noise level, and information retention of each high-frequency sub-band component. The denoised ultrasonic data is then matched with a pre-established fatigue damage database and a non-damage database. The results of fatigue damage detection of micro-precision bearings are obtained based on the matching results.

[0005] Furthermore, the specific calculation method for the initial threshold of the high-frequency sub-band component is as follows: The length of the high-frequency sub-band component is taken as the argument of the logarithm with the natural constant as the base. The product of the square root of twice the value of the logarithm and the standard deviation of all the data contained in the high-frequency sub-band component is denoted as the initial threshold of the high-frequency sub-band component.

[0006] Furthermore, the specific method for obtaining the interference level of the high-frequency sub-band components is as follows: The normalized value of the variance of all data contained in the high-frequency sub-band component is denoted as the interference degree of the high-frequency sub-band component.

[0007] Furthermore, the specific calculation method for the non-smoothness of the high-frequency sub-band components is as follows: The upper and lower envelope curves are smoothed separately. Based on the differences between the upper and lower envelope curves before and after smoothing, the smoothing differences between the upper and lower envelope curves are calculated. The mean of the smoothness difference between the upper and lower envelope curves is denoted as the unsmoothness of the high-frequency sub-band component.

[0008] Furthermore, the specific calculation method for the smoothing difference between the upper and lower envelope curves is as follows: The normalized value of the DTW distance between the upper envelope curves before and after smoothing is denoted as the smoothing difference of the upper envelope curve; the normalized value of the DTW distance between the lower envelope curves before and after smoothing is denoted as the smoothing difference of the lower envelope curve.

[0009] Furthermore, the method for determining the noise content of the high-frequency sub-band components is as follows: The product of the interference level and the unsmoothness of the high-frequency sub-band component is denoted as the noise level of the high-frequency sub-band component.

[0010] Furthermore, the method for calculating the information similarity of the target high-frequency sub-band components is as follows: The reconstructed ultrasonic data of all high-frequency sub-band components except the target high-frequency sub-band component is denoted as the reconstructed ultrasonic data. The normalized value of the DTW distance between the reconstructed ultrasonic data and the acquired ultrasonic data is denoted as the missing difference of the target high-frequency sub-band component. The difference between the number 1 and the missing difference of the target high-frequency sub-band component is denoted as the information similarity of the target high-frequency sub-band component.

[0011] Furthermore, the specific method for obtaining the information retention degree of each high-frequency sub-band component based on the number of clusters obtained by clustering and the difference in information similarity within different clusters includes: When the number of clusters obtained by clustering is not 2, the information retention level of all high-frequency sub-band components is assigned to 1. When the number of clusters obtained by clustering is 2, the normalized value of the difference between the mean of all information similarity in the cluster where the target high-frequency sub-band component is located and the mean of all information similarity in another cluster is recorded as the information retention degree of the target high-frequency sub-band component.

[0012] Furthermore, the specific method for denoising the ultrasonic data based on the initial threshold, noise level, and information retention degree of each high-frequency sub-band component includes: The ratio of the product of the initial threshold of the high-frequency sub-band component and the noise level to the degree of information preservation is denoted as the correction threshold of the high-frequency sub-band component. The correction threshold of the high-frequency sub-band component is used as the wavelet threshold corresponding to the high-frequency sub-band component, and the wavelet threshold denoising algorithm is used to denoise the ultrasonic data.

[0013] Secondly, embodiments of the present invention also provide a micro-precision bearing fatigue damage detection system, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0014] The beneficial effects of this invention are: This application first decomposes high-frequency sub-band components from ultrasonic data and calculates initial thresholds based on these components. When the high-frequency sub-band components are more affected by noise and retain more information from the ultrasonic data, the less effective information is retained within them. Therefore, more precise suppression and removal of noise information within the high-frequency sub-band components is necessary. The smoothness and value differences of the high-frequency sub-band components are analyzed to evaluate the degree of noise interference and obtain their noise content. Furthermore, ultrasonic data is reconstructed for all high-frequency sub-band components except the target high-frequency sub-band component. Based on the difference between the reconstructed ultrasonic data and the acquired ultrasonic data, the information similarity of the target high-frequency sub-band component is calculated, and the reconstruction is completed. The greater the difference between the ultrasonic data and the acquired ultrasonic data, the more completely the target high-frequency sub-band component retains information from the acquired ultrasonic data. In this case, the information similarity of the target high-frequency sub-band component is smaller. Based on the number of clusters obtained through clustering and the difference in information similarity within different clusters, the information retention degree of each high-frequency sub-band component is obtained. Finally, a wavelet threshold denoising algorithm is used to denoise the ultrasonic data based on the initial threshold, noise level, and information retention degree of each high-frequency sub-band component. Based on the denoised ultrasonic data, the results of fatigue damage detection for micro-precision bearings are obtained, solving the problem of inaccurate fatigue damage detection results due to poor denoising of ultrasonic data for micro-precision bearings, and improving the accuracy of fatigue damage detection for micro-precision bearings. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating a method for detecting fatigue damage in miniature precision bearings according to an embodiment of the present invention. Figure 2 This is a flowchart of the initial threshold acquisition process provided in one embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 The diagram illustrates a flowchart of a method for detecting fatigue damage in micro-precision bearings according to an embodiment of the present invention. The method includes the following steps: Step S001: Collect ultrasonic data of the micro-precision bearing, decompose the ultrasonic data into low-frequency sub-band components and different high-frequency sub-band components, and calculate the initial threshold of the high-frequency sub-band components.

[0019] Ultrasonic data are collected from tiny precision bearings to be tested for fatigue damage.

[0020] Wavelet decomposition is used to decompose the ultrasonic data into a first preset number of high-frequency sub-band components and one low-frequency sub-band component.

[0021] In this embodiment, the first preset quantity is set to 5; when performing wavelet decomposition, the DB8 wavelet basis function is used.

[0022] The initial threshold of the high-frequency sub-band component is calculated based on the length of the high-frequency sub-band component and the standard deviation of all data contained in the high-frequency sub-band component.

[0023] Preferably, as an embodiment of this application, the length of the high-frequency sub-band component is taken as the argument of the logarithm with the natural constant as the base, and the product of the square root of twice the logarithm value and the standard deviation of all data contained in the high-frequency sub-band component is recorded as the initial threshold of the high-frequency sub-band component.

[0024] At this point, the initial threshold for each high-frequency sub-band component is obtained. The flowchart for obtaining the initial threshold is as follows: Figure 2 As shown.

[0025] Step S002: Determine the interference degree of the high-frequency sub-band component based on the differences in all the data contained in the high-frequency sub-band component; calculate the non-smoothness of the high-frequency sub-band component based on the differences in the upper and lower envelope curves of the high-frequency sub-band component before and after smoothing; and determine the noise content of the high-frequency sub-band component based on the interference degree and non-smoothness of the high-frequency sub-band component.

[0026] The greater the impact of noise on the high-frequency sub-band components and the greater the degree of information retention in the ultrasonic data, the less effective information is retained within the high-frequency sub-band components. Therefore, it is necessary to more precisely suppress and eliminate noise information within the high-frequency sub-band components.

[0027] The interference level of the high-frequency sub-band component is determined based on the differences in all the data contained in the high-frequency sub-band component.

[0028] The normalized value of the variance of all data contained in the high-frequency sub-band component is denoted as the interference degree of the high-frequency sub-band component.

[0029] In this embodiment, the sigmoid function is used to calculate the normalized value. The sigmoid function is a well-known technique and will not be described in detail here. As other implementation methods, implementers can use other methods of the prior art, such as the tanh function.

[0030] The greater the interference of the high-frequency sub-band components, the more noise interference the high-frequency sub-band components will experience.

[0031] Noise can disrupt the accuracy of local extrema in high-frequency sub-band components, causing jitter or distortion in the envelope, which manifests as reduced envelope smoothness. Therefore, the degree of noise interference on high-frequency sub-band components can be evaluated based on the smoothness of the high-frequency sub-band components.

[0032] The upper and lower envelope curves of the high-frequency sub-band components are obtained separately. Based on the difference between the upper and lower envelope curves of the high-frequency sub-band components before and after smoothing, the non-smoothness of the high-frequency sub-band components is calculated.

[0033] Preferably, as an embodiment of this application, mean filtering is used to smooth the upper envelope curve and the lower envelope curve respectively. The normalized value of the DTW distance of the upper envelope curve before and after smoothing is recorded as the smoothing difference of the upper envelope curve, and the normalized value of the DTW distance of the lower envelope curve before and after smoothing is recorded as the smoothing difference of the lower envelope curve. The mean of the smoothing differences of the upper envelope curve and the lower envelope curve is recorded as the non-smoothness of the high-frequency sub-band component.

[0034] In this embodiment, the step size of the mean filtering is set to 3. The use of mean envelope curve filtering to smooth the curve is a well-known technique and will not be described in detail here.

[0035] The result of the positive correlation processing between the interference degree and the unsmoothness of the high-frequency sub-band components is denoted as the noise content of the high-frequency sub-band components.

[0036] It is understood that a positive correlation processing is applied to the interference and non-smoothness of the high-frequency sub-band components, ensuring that the interference and non-smoothness of the high-frequency sub-band components are positively correlated with the noise content of the high-frequency sub-band components. It is also understood that the positive correlation in this application refers to the relationship between the independent and dependent variables, where the independent variables are the interference and non-smoothness of the high-frequency sub-band components, and the dependent variable is the noise content of the high-frequency sub-band components. The positive correlation means that the dependent variable increases (decreases) as the independent variable increases (decreases), and can be an additive or multiplicative relationship.

[0037] Preferably, as an embodiment of this application, the product of the interference degree and the unsmoothness of the high-frequency sub-band component is denoted as the noise content of the high-frequency sub-band component.

[0038] At this point, the noise content of each high-frequency sub-band component is obtained.

[0039] Step S003: Denote any high-frequency sub-band component as the target high-frequency sub-band component. Based on the difference between the reconstruction results of the ultrasonic data of all high-frequency sub-band components other than the target high-frequency sub-band component and the acquired ultrasonic data, calculate the information similarity of the target high-frequency sub-band component. Cluster the information similarity of all high-frequency sub-band components. Based on the number of clusters obtained by clustering and the difference in information similarity within different clusters, obtain the information retention degree of each high-frequency sub-band component.

[0040] Any high-frequency sub-band component is designated as the target high-frequency sub-band component. Ultrasonic data is reconstructed for all high-frequency sub-band components except the target high-frequency sub-band component. The greater the difference between the reconstructed ultrasonic data and the acquired ultrasonic data, the more completely the target high-frequency sub-band component retains the information in the acquired ultrasonic data.

[0041] Reconstruct ultrasonic data for all high-frequency sub-band components except the target high-frequency sub-band component, obtain reconstructed ultrasonic data, and record the normalized value of the DTW distance between the reconstructed ultrasonic data and the acquired ultrasonic data as the missing difference of the target high-frequency sub-band component. Record the difference between the number 1 and the missing difference of the target high-frequency sub-band component as the information similarity of the target high-frequency sub-band component.

[0042] The smaller the information similarity of the target high-frequency sub-band components, the more completely the target high-frequency sub-band components retain information in the acquired ultrasonic data.

[0043] The information similarity of any high-frequency sub-band component can be obtained using the same method.

[0044] Cluster the information similarity of all high-frequency sub-band components and count the number of clusters obtained from the clustering.

[0045] In this embodiment, the AP clustering algorithm is used to achieve clustering. In practical applications, in addition to achieving the goal of clustering, implementers may use other existing methods such as DBSCAN clustering algorithm, Mean Shift clustering algorithm, OPTICS clustering algorithm, Gaussian Mixture Models clustering algorithm, and Spectral Clustering for clustering. This application does not impose any special restrictions.

[0046] When the number of clusters obtained by clustering is not 2, the information retention level of all high-frequency sub-band components is assigned to 1.

[0047] When the number of clusters obtained by clustering is 2, in these two clusters, the high-frequency sub-band components corresponding to the information similarity contained in one cluster retain the information in the acquired ultrasonic data more completely, while the high-frequency sub-band components corresponding to the information similarity contained in the other cluster retain the information in the acquired ultrasonic data less completely. Based on the difference in information similarity within different clusters, the information retention degree of the target high-frequency sub-band component is determined. Preferably, as an embodiment of this application, the normalized value of the difference between the mean of all information similarities in the cluster where the target high-frequency sub-band component is located and the mean of all information similarities in the other cluster is recorded as the information retention degree of the target high-frequency sub-band component.

[0048] The same method can be used to obtain the information retention level of any high-frequency sub-band component.

[0049] At this point, the degree of information retention for each high-frequency sub-band component is obtained.

[0050] Step S004: Using a wavelet threshold denoising algorithm, the ultrasonic data is denoised based on the initial threshold, noise level, and information retention of each high-frequency sub-band component to obtain denoised ultrasonic data. The denoised ultrasonic data is then matched with a pre-established fatigue damage database and a non-damage database. The results of fatigue damage detection of micro-precision bearings are obtained based on the matching results.

[0051] The correction threshold of the high-frequency sub-band component is calculated based on the initial threshold, noise content, and information retention of the high-frequency sub-band component.

[0052] Preferably, as an embodiment of this application, the ratio of the product of the initial threshold of the high-frequency sub-band component and the noise level to the information retention level is denoted as the correction threshold of the high-frequency sub-band component.

[0053] It should be noted that in the process of calculating the ratio, in order to avoid the denominator being zero, a preset value needs to be added to the denominator. In this example, the preset value is 0.01.

[0054] When the noise content of the high-frequency sub-band components is greater and the information retention is smaller, a larger threshold should be assigned to the high-frequency sub-band components when using wavelet threshold denoising algorithm to denoise ultrasonic data, so that the high-frequency sub-band components are easier to remove, thereby reducing the interference of noise on ultrasonic data.

[0055] The correction threshold of the high-frequency sub-band component is used as the wavelet threshold corresponding to the high-frequency sub-band component. The wavelet threshold denoising algorithm is used to denoise the ultrasonic data to obtain the denoised ultrasonic data.

[0056] Experts in the field established fatigue damage and non-damage databases for miniature precision bearings, composed of ultrasonic data. Noise-reduced ultrasonic data was then matched against the ultrasonic data in both databases to obtain the optimal matching result. If the optimal matching result corresponds to the fatigue damage database, the miniature precision bearing to be tested for fatigue damage is determined to have fatigue damage; if the optimal matching result corresponds to the non-damage database, the miniature precision bearing to be tested for fatigue damage is determined to not have fatigue damage.

[0057] Among them, establishing fatigue damage databases and non-damage databases for micro-precision bearings and matching ultrasonic data are well-known technologies and will not be elaborated further.

[0058] This enables the detection of fatigue damage in minute precision bearings.

[0059] Based on the same inventive concept as the above method, this embodiment of the invention also provides a micro-precision bearing fatigue damage detection system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described micro-precision bearing fatigue damage detection methods.

[0060] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for detecting fatigue damage in micro-precision bearings, characterized in that, The method includes the following steps: Ultrasonic data of micro-precision bearings are collected, and the ultrasonic data is decomposed into low-frequency sub-band components and different high-frequency sub-band components. The initial threshold of the high-frequency sub-band components is calculated. Based on the differences in all the data contained in the high-frequency sub-band component, the interference degree of the high-frequency sub-band component is determined. Based on the differences in the upper and lower envelope curves of the high-frequency sub-band component before and after smoothing, the non-smoothness of the high-frequency sub-band component is calculated. Based on the interference degree and non-smoothness of the high-frequency sub-band component, the noise content of the high-frequency sub-band component is determined. Any high-frequency sub-band component is denoted as the target high-frequency sub-band component. Based on the difference between the reconstruction results of the ultrasonic data of all high-frequency sub-band components other than the target high-frequency sub-band component and the acquired ultrasonic data, the information similarity of the target high-frequency sub-band component is calculated. The information similarity of all high-frequency sub-band components is clustered. Based on the number of clusters obtained by clustering and the difference in information similarity within different clusters, the information retention degree of each high-frequency sub-band component is obtained. The wavelet threshold denoising algorithm is used to denoise the ultrasonic data according to the initial threshold, noise content and information retention of each high-frequency sub-band component, and the denoised ultrasonic data is obtained. The denoised ultrasonic data is then matched with a pre-established fatigue damage database and a non-damage database. The results of fatigue damage detection of micro-precision bearings are obtained based on the matching results. The upper and lower envelope curves are smoothed separately. Based on the difference between the upper and lower envelope curves before and after smoothing, the smoothing difference between the upper and lower envelope curves is calculated. The mean of the smoothing difference between the upper and lower envelope curves is denoted as the non-smoothness of the high-frequency sub-band component. The normalized value of the DTW distance between the upper envelope curves before and after smoothing is denoted as the smoothing difference of the upper envelope curve; the normalized value of the DTW distance between the lower envelope curves before and after smoothing is denoted as the smoothing difference of the lower envelope curve. The calculation method for the information similarity of the target high-frequency sub-band components is as follows: the reconstruction results of the ultrasonic data of all high-frequency sub-band components other than the target high-frequency sub-band components are denoted as the reconstructed ultrasonic data; the normalized value of the DTW distance between the reconstructed ultrasonic data and the acquired ultrasonic data is denoted as the missing difference of the target high-frequency sub-band components; and the difference between the number 1 and the missing difference of the target high-frequency sub-band components is denoted as the information similarity of the target high-frequency sub-band components. Specifically, when the number of clusters obtained by clustering is not 2, the information retention degree of all high-frequency sub-band components is assigned to 1; when the number of clusters obtained by clustering is 2, the normalized value of the difference between the mean of all information similarity in the cluster where the target high-frequency sub-band component is located and the mean of all information similarity in another cluster is recorded as the information retention degree of the target high-frequency sub-band component. The product of the initial threshold and the noise level of the high-frequency sub-band component is denoted as the ratio of the information retention level to the correction threshold of the high-frequency sub-band component. The correction threshold of the high-frequency sub-band component is used as the wavelet threshold corresponding to the high-frequency sub-band component, and the wavelet threshold denoising algorithm is used to denoise the ultrasonic data.

2. The method for detecting fatigue damage in micro-precision bearings according to claim 1, characterized in that, The specific calculation method for the initial threshold of the high-frequency sub-band component is as follows: The length of the high-frequency sub-band component is taken as the argument of the logarithm with the natural constant as the base. The product of the square root of twice the value of the logarithm and the standard deviation of all the data contained in the high-frequency sub-band component is denoted as the initial threshold of the high-frequency sub-band component.

3. The method for detecting fatigue damage in micro-precision bearings according to claim 1, characterized in that, The specific method for obtaining the interference level of the high-frequency sub-band components is as follows: The normalized value of the variance of all data contained in the high-frequency sub-band component is denoted as the interference degree of the high-frequency sub-band component.

4. The method for detecting fatigue damage in micro-precision bearings according to claim 1, characterized in that, The method for determining the noise content of the high-frequency sub-band components is as follows: The product of the interference level and the unsmoothness of the high-frequency sub-band component is denoted as the noise level of the high-frequency sub-band component.

5. A micro-precision bearing fatigue damage detection system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as claimed in any one of claims 1-4.

Images