Rolling bearing fault diagnosis method under variable speed working condition

Abstract

The invention discloses a rolling bearing fault diagnosis method under a variable speed working condition. The method comprises the following steps: firstly, performing low-pass filtering and Hilbert transform on a variable-speed rolling bearing vibration signal to obtain an envelope signal, performing angle domain resampling on the envelope signal through a calculation order tracking method (COT), and converting a non-stable vibration signal under a variable-speed working condition into a relatively stable vibration signal under an angle domain; secondly, decomposing the angle domain vibration signal by utilizing inherent time scale decomposition (ITD) to obtain a plurality of inherent rotation components and a monotonic trend term; and finally, by calculating the kurtosis and the correlation coefficient of each component, screening a proper component for reconstruction, and performing order analysis on a reconstructed signal to complete fault diagnosis of the variable-speed rolling bearing. The COT and ITD combined method can effectively extract the fault feature order of the rolling bearing under noise interference, has a good diagnosis effect on the rolling bearing fault under the variable speed working condition, and has practical application value.

Description

Technical Field

[0001] The invention relates to the field of rotating machinery fault diagnosis, and in particular to a fault diagnosis method when a rolling bearing fails under variable speed conditions. Background Art

[0002] As a key component of rotating mechanical equipment, rolling bearings are widely used in aviation and navigation, rail trains, automobiles, mechanical processing, wind turbines and other fields. When a rolling bearing fails, it may trigger a chain reaction that causes the entire mechanical system to fail, affecting the normal operation of the equipment and causing serious economic losses.

[0003] When a rolling bearing fails, the vibration signal it generates contains important fault feature information. Common rolling bearing failures include inner ring failure, outer ring failure, rolling element failure, etc. Under constant speed conditions, time-frequency domain analysis methods are usually used to extract fault features from rolling bearing vibration signals to determine the type of bearing failure. However, in actual industrial production processes, the operating state of rolling bearings is complex and changeable. Under variable speed conditions, the fault feature frequency will change due to the change in speed. At this time, the vibration signal is a nonlinear and non-stationary signal. Only through time-frequency domain analysis, "spectral blurring" will occur, making it difficult to effectively extract the fault features contained in the rolling bearing vibration signal. In order to deal with the problem of nonlinear and non-stationary variable speed rolling bearing vibration signal processing, experts and scholars have proposed an order analysis method, which is widely used. Hardware order tracking uses hardware to perform equal-angle resampling to obtain a smooth signal in the angular domain. However, due to the hardware cost and installation location to be considered in actual use, there are certain limitations. The computed order tracking method (Computed Order Tracking, COT) achieves equal-angle resampling through software methods. While reducing the required hardware, it can ensure the accuracy of angle resampling, so it has important practical application value.

[0004] The above-mentioned calculation order tracking method can convert the nonlinear and non-stationary variable speed rolling bearing fault signal into a relatively stable variable speed vibration signal in the angle domain through angle resampling. However, in the actual industrial environment, the collected rolling bearing vibration signal, its fault characteristics are easily affected by noise interference. Even if the stable signal in the angle domain is obtained by the angle resampling method, it is still an FM signal with noise interference. Therefore, it is necessary to take noise reduction processing for the vibration signal in the angle domain, highlight the fault characteristics of the vibration signal, and then perform order analysis on the denoised angle domain signal, so that it is possible to extract the fault characteristics of the rolling bearing under variable speed working conditions under noise interference, thereby achieving the purpose of fault diagnosis.

[0005] Frei et al. proposed the intrinsic time-scale decomposition (ITD) method, which provides a new idea for signal decomposition. ITD can decompose the vibration signal into multiple intrinsic rotation components (ProperRotation, PR) and a monotonic trend term. Compared with the empirical mode decomposition (EMD) method, the ITD method reduces the computational complexity, improves the computational efficiency, and can avoid the problems of the EMD method such as the endpoint effect to a certain extent. Based on the high computational efficiency of ITD and its advantages over EMD, ITD has been widely used in the field of rotating machinery fault diagnosis.

[0006] According to the above content, in view of the fact that the fault vibration signal of the rolling bearing under variable speed conditions is easily affected by noise interference and it is difficult to extract the fault characteristics, a method combining COT calculation order tracking and ITD intrinsic time modal decomposition is proposed to solve this problem. First, the rolling bearing vibration signal under variable speed is low-pass filtered and Hilbert transformed, and then the angle domain is resampled to obtain a relatively stable angle domain signal. After using ITD to decompose the angle domain signal, a number of PR components are generated. The correlation coefficient and kurtosis are used as the selection indexes of these PR components, and the appropriate PR components are screened out and the signal is reconstructed. The reconstructed signal is subjected to order analysis, and the fault diagnosis is completed according to the order analysis results. The present invention can effectively extract the rolling bearing fault characteristics from the vibration signal containing noise and other interferences under variable speed conditions, has a good fault diagnosis effect, and has practical application value. Summary of the invention

[0007] The present invention proposes a rolling bearing fault diagnosis method under variable speed conditions based on a combination of angle resampling and inherent time decomposition, which is used to solve the problem that under variable speed conditions, bearing faults are affected by speed changes and noise interference, and fault features are difficult to effectively extract, resulting in difficulty in fault diagnosis.

[0008] In order to achieve the above-mentioned invention object, the specific technical scheme proposed is as follows: S1. Low-pass filtering is performed on the bearing vibration acceleration signal. The filtered signal is subjected to Hilbert transform to obtain an envelope signal, and then angle domain resampling is performed to obtain a vibration signal in the angle domain. S2. Decompose the vibration signal in the angle domain by ITD to obtain several PR components; S3, respectively calculating the correlation coefficient between each PR component and the angle domain signal before ITD decomposition, and then calculating the kurtosis value of each PR component; S4, comprehensively screening the PR components with larger correlation coefficients and kurtosis indexes to reconstruct the signal, and analyzing the reconstructed signal through order envelope spectrum to realize rolling bearing fault diagnosis under variable speed conditions;

[0009] Furthermore, the specific contents of the above specific technical solution S1 include: Set the low-pass filter parameters to reduce the interference of medium and high frequency noise caused by variable speed working conditions, and highlight the low-frequency fault characteristics of the bearing; Perform Hilbert transform on the filtered signal to obtain the envelope signal; The time of equal-angle sampling is calculated by collecting the speed pulse signal, and the angle resampling is realized by adopting the piecewise linear interpolation method.

[0010] Furthermore, the specific contents of the above-mentioned specific technical solution S1 include: the specific steps of implementing equal angle resampling are as follows: assuming that the axis runs in a uniformly accelerated state, the angle of rotation is and time taken It can be expressed by the following equation: , where , , are the polynomial coefficients, t\left ( {\theta} \right )=\frac {1} {2{a}_{2}}\left [ {\sqrt {4{a}_{2\left ( {} \right )}\left ( {\theta -{a}_{0}} \right )+{a}^{2}_{1}}-{a}_{1}} \right ] , usually, the resampling process is discrete and can be expressed as: , represents the expected angular interval between resampling, then the expressions for angle and time can be expressed as: t\left ( {k\Delta \theta} \right )=\frac {1} {2{a}_{2}}\left [ {\sqrt {4{a}_{2\left ( {} \right )}\left ( {k\Delta \theta -{a}_{0}} \right )+{a}^{2}_{1}}-{a}_{1}} \right ] The above expression can be used to calculate the angle resampling of order tracking.

[0011] The rolling bearing fault diagnosis method under variable speed working condition according to claim 1 is characterized in that: in step S2, the specific processing process is as follows: after using ITD to decompose the angle domain signal, a plurality of PR components are generated. The steps include: ITD can convert bearing vibration signals Decomposed into a set of intrinsic rotation components and a monotonic trend baseline component . It can be expressed as: ,in Extract the factors for the baseline and solve The process is as follows, where for The extreme point of , on the interval of consecutive extreme points\left [ {{\tau}_{k},{\tau}_{k+1}} \right ] Extraction factor can be determined , , {L}_{k+1}=\alpha \left [ {{X}_{k}+\frac {{\tau}_{k+1}-{\tau}_{k}} {{\tau}_{k+2}-{\tau}_{k}}\left ( {{X}_{k+2}-{X}_{k}} \right )} \right ]+\left ( {1-\alpha} \right ){X}_{k+1} , where is the linear scaling factor, which is used to adjust the linear scaling of the vibration component amplitude. It is usually set to 0.5. Seen as vibration signal Repeat the above decomposition process. When it is a monotonic trend signal, the entire decomposition process can be expressed as .

[0012] The rolling bearing fault diagnosis method under variable speed conditions according to claim 1 is characterized in that: in step S3, the specific processing process is as follows: The correlation coefficient and kurtosis value of the PR components generated by ITD decomposition are calculated. After comprehensive comparison, the PR components with large correlation coefficient and kurtosis are selected, and the signals of the selected PR components are reconstructed.

[0013] The rolling bearing fault diagnosis method under variable speed conditions according to claim 1 is characterized in that: in step S4, the specific processing process is as follows: The reconstructed signal is subjected to order analysis after FFT transformation, and the specific order of the bearing fault is obtained according to the order envelope spectrum. By comparing and analyzing it with the calculated theoretical order of rolling bearing fault, it is determined which specific rolling bearing fault the current rolling bearing fault belongs to, thereby realizing rolling bearing fault diagnosis under variable speed conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 The schematic diagram of the rolling bearing fault diagnosis process under the variable speed working condition implemented in this application is listed; Figure 2 It is the time domain diagram of the vibration signal under deceleration condition; Figure 3 It is the order spectrum with noise interference after preprocessing and angle resampling; Figure 4 The first five PR components after ITD decomposition of the noisy angle domain vibration signal; Figure 5 Order envelope spectrum for order analysis of the reconstructed signal; DETAILED DESCRIPTION

[0015] To make the above-mentioned purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the embodiments described in the present invention are only used to explain the present invention, not to limit the present invention.

[0016] Please refer to Figure 1-Figure 5 The accompanying drawings show a specific implementation method. This implementation combines the calculation order tracking method and the ITD decomposition method. The method includes first filtering the original vibration signal and Hilbert transform processing, and resampling the processed signal in the angle domain. The purpose is to convert the nonlinear and non-stationary vibration signal into a relatively stable vibration signal in the angle domain. The vibration signal in the angle domain also contains noise signals and other noise signals at this time. Next, the vibration signal in the angle domain is decomposed by the ITD decomposition method to generate several PR components. According to the correlation coefficient and kurtosis criterion, the correlation coefficient and kurtosis of each PR component are calculated, and the appropriate PR component is selected for signal reconstruction. The above-mentioned ITD processing and signal reconstruction can reduce the influence of noise and other interference signals in the original angle domain signal, and improve the ability to extract rolling bearing fault features under variable speed conditions.

[0017] In step S1, the bearing vibration acceleration signal is subjected to low-pass filtering, the filtered signal is subjected to Hilbert transform to obtain the envelope signal, and then the angle domain is resampled to obtain the vibration signal in the angle domain.

[0018] In step S2, the vibration signal in the angle domain is decomposed by ITD to obtain several PR components. Compared with EMD empirical mode decomposition, ITD intrinsic time scale decomposition reduces the computational complexity, improves the computational efficiency, and can avoid problems such as endpoint effects to a certain extent.

[0019] In step S3, the correlation coefficient between each PR component and the angle domain signal before ITD decomposition is calculated respectively, and then the kurtosis value of each PR component is calculated.

[0020] In step S4, PR components with larger correlation coefficients and kurtosis indexes are comprehensively screened for signal reconstruction, and the reconstructed signals are subjected to order analysis to realize rolling bearing fault feature extraction under variable speed conditions.

[0021] In order to verify the fault diagnosis effect of the present invention, the experimental data is analyzed using the variable speed bearing vibration signal data set of the University of Ottawa, Canada. From the data set parameters, it can be seen that the characteristic order of the inner ring fault of the rolling bearing in this data set is 5.43. The vibration signal of the inner ring fault of the rolling bearing is as follows under the deceleration condition: Figure 2 shown.

[0022] The envelope spectrum analysis of the rolling bearing vibration signal shows that due to the change in speed, the "spectrum blurring" phenomenon occurs, and it is difficult to directly obtain the fault characteristics of the rolling bearing under variable speed conditions through frequency domain analysis.

[0023] Furthermore, the vibration signal is angle resampled, and the specific steps of angle resampling include: The specific steps to achieve equal angle resampling are as follows: Assume that the axis is running at a uniform acceleration, and its rotation angle is and time taken It can be expressed by the following equation: , where , , are the polynomial coefficients, t\left ( {\theta} \right )=\frac {1} {2{a}_{2}}\left [ {\sqrt {4{a}_{2\left ( {} \right )}\left ( {\theta -{a}_{0}} \right )+{a}^{2}_{1}}-{a}_{1}} \right ] , usually, the resampling process is discrete and can be expressed as: , represents the expected angular interval between resampling, then the expressions for angle and time can be expressed as: t\left ( {k\Delta \theta} \right )=\frac {1} {2{a}_{2}}\left [ {\sqrt {4{a}_{2\left ( {} \right )}\left ( {k\Delta \theta -{a}_{0}} \right )+{a}^{2}_{1}}-{a}_{1}} \right ] The above expression can be used to calculate the angle resampling of order tracking.

[0024] Through angle resampling, the nonlinear and non-stationary vibration signal is converted into a vibration signal in the angle domain.

[0025] Furthermore, since the converted angle domain signal still contains noise interference and other noise interference signals, the order envelope spectrum is as follows: Figure 3 shown.

[0026] Furthermore, the converted angle domain signal is subjected to ITD decomposition.

[0027] In step S2, the vibration signal in the angle domain is decomposed by ITD. ITD can decompose the bearing vibration signal Decomposed into a set of intrinsic rotation components and a monotonic trend baseline component . It can be expressed as: ,in Extract the factors for the baseline and solve The process is as follows, where for The extreme point of , on the interval of consecutive extreme points\left [ {{\tau}_{k},{\tau}_{k+1}} \right ] Extraction factor can be determined , , {L}_{k+1}=\alpha \left [ {{X}_{k}+\frac {{\tau}_{k+1}-{\tau}_{k}} {{\tau}_{k+2}-{\tau}_{k}}\left ( {{X}_{k+2}-{X}_{k}} \right )} \right ]+\left ( {1-\alpha} \right ){X}_{k+1} , where is the linear scaling factor, which is used to adjust the linear scaling of the vibration component amplitude. It is usually set to 0.5. Seen as vibration signal Repeat the above decomposition process. When it is a monotonic trend signal, the entire decomposition process can be expressed as After decomposition, several PR components are generated. The first five PR components after decomposition are as follows: Figure 4 shown.

[0028] The step according to claim 1 is characterized in that: in step S3, the specific processing process is as follows: The correlation coefficient and kurtosis value of the PR components generated by ITD decomposition are calculated. After comprehensive comparison, the PR components with large correlation coefficient and kurtosis are selected, and the signals of the selected PR components are reconstructed.

[0029] The order analysis of the reconstructed signal is performed, and the specific order of the bearing fault is obtained according to the order spectrum, thus completing the extraction of the characteristic order of the rolling bearing fault.

[0030] The specific order of rolling bearing fault obtained through the order envelope spectrum is compared and analyzed with the theoretical order of rolling bearing fault obtained by calculation to determine which specific fault the current rolling bearing fault order belongs to, thereby achieving the ultimate goal of rolling bearing fault diagnosis under variable speed conditions.

[0031] Figure 5 This is the spectrum of the order analysis of the reconstructed signal using the COT combined with ITD method. Figure 5 The fault order can be obtained from 5.41, 2 times the fault order 10.81 and 3 times the fault order is 16.21. By comparing with the theoretical fault feature order of the data set, the actual fault order is close to the theoretical fault order of 5.43, and the bearing fault type is judged to be an inner ring fault. And by comparing with the order spectrum without ITD processing, the proposed method can more intuitively obtain the fault order of the rolling bearing under variable speed conditions and reduce the interference of other feature orders.

[0032] Under variable speed conditions, the non-stationary vibration signal of the rolling bearing is converted into a relatively stable signal in the angular domain by calculating the order tracking method, thus avoiding the spectrum ambiguity problem caused by speed changes. The order spectrum obtained by the method proposed in the present invention can more intuitively reflect the fault characteristic order. By calculating the kurtosis and correlation coefficient, the appropriate PR component is selected for reconstruction, reducing the work of manual component selection. The reconstructed angular domain signal can effectively extract the rolling bearing fault characteristic order from noise interference through order analysis, achieving good fault diagnosis effect.

Claims

1. The present invention proposes a rolling bearing fault diagnosis method under variable speed conditions. It is characterized by: The following steps are included S1, low-pass filtering is performed on the bearing vibration acceleration signal, the filtered signal is subjected to Hilbert transform to obtain the envelope signal, and then angle domain resampling is performed to obtain the vibration signal in the angle domain; S2, decompose the vibration signal in the angle domain by the intrinsic time scale decomposition method ITD to obtain several PR components; S3, respectively calculating the correlation coefficient between each PR component and the angle domain signal before ITD decomposition, and then calculating the kurtosis value of each PR component; S4, select the PR components with larger correlation coefficient and kurtosis index to reconstruct the signal, and perform order analysis on the reconstructed signal to realize the extraction of rolling bearing fault features under variable speed conditions.

2. The rolling bearing fault diagnosis method under variable speed conditions according to claim 1 is characterized in that: In step S1, the specific processing process is as follows: S11, setting low-pass filter parameters, using low-pass filters to reduce the interference of medium and high frequency noises caused by variable speed working conditions, and enhancing the characteristics of bearing faults; S12, performing Hilbert transform on the filtered signal to obtain an envelope signal; S13, calculating the time of equal-angle sampling through the collected speed pulse signal, and implementing angle resampling by adopting a piecewise linear interpolation method; S14, the specific steps to achieve equal angle resampling are as follows: Assume that the axis is running under uniform acceleration, and its rotation angle and time taken It can be expressed by the following equation: , where , , are the polynomial coefficients, , usually, the resampling process is discrete and can be expressed as: , represents the desired angular interval between resampling, then the expressions for angle and time can be expressed as: The above expression can be used to calculate the angle resampling of order tracking.

3. The rolling bearing fault diagnosis method under variable speed conditions according to claim 1 is characterized in that: In step S2, the specific processing process is as follows: After using ITD to decompose the angle domain signal, several PR components are generated. The characteristics are: ITD can convert bearing vibration signals Decomposed into a set of intrinsic rotation components and a monotonic trend baseline component . It can be expressed as: ,in Extract the factors for the baseline and solve The process is as follows, where for The extreme point of , on the intervals between consecutive extreme points Extraction factor can be determined , , , where is the linear scaling factor, which is used to adjust the linear scaling of the vibration component amplitude. It is usually set to 0.

5. Seen as vibration signal Repeat the above decomposition process. When it is a monotonic trend signal, the entire decomposition process can be expressed as .

4. The rolling bearing fault diagnosis method under variable speed conditions according to claim 1 is characterized in that: In step S3, the specific processing process is as follows: calculate the correlation coefficient and kurtosis value of the PR components generated by ITD decomposition, comprehensively compare and select PR components with large correlation coefficients and kurtosis, and then reconstruct the signals of the selected PR components.

5. The rolling bearing fault diagnosis method under variable speed conditions according to claim 1 is characterized in that: In step S4, the specific processing process is as follows: perform order analysis on the reconstructed signal after FFT transformation, obtain the bearing fault characteristic order according to the order envelope spectrum, complete the extraction of the rolling bearing fault characteristic order, and realize the rolling bearing fault diagnosis under variable speed conditions by comparing and judging the fault characteristics with the theoretical calculation.

Images