Railway wagon bearing fault diagnosis method based on EEMD and BWO-VMD

Abstract

The invention provides a railway wagon bearing fault diagnosis method based on EEMD and BWO-VMD, and belongs to the technical field of railway safety detection. An original vibration signal of a truck bearing is collected; reconstructing an original signal through EEMD; decomposing the reconstructed signal through BWO-VMD to obtain a plurality of IMF components; performing envelope demodulation by using the non-aliasing IMF components to obtain an envelope spectrum frequency diagram; and performing fault diagnosis by matching the envelope spectrum frequency diagram with the bearing fault characteristic frequency. The endpoint effect and the mode aliasing phenomenon existing in the railway wagon bearing vibration signal are improved, and therefore the accuracy of signal extraction is improved.

Description

Technical Field

[0001] The invention relates to the technical field of railway safety detection, and in particular to a railway freight car bearing fault diagnosis method based on EEMD and BWO-VMD. Background Art

[0002] Railway bearings are one of the key components of railway freight cars, and their condition is closely related to the safety of railway freight car operation. Therefore, being able to accurately separate and identify railway freight car bearing fault information under noise interference is of great significance to ensuring the safe operation of railway freight cars, reducing maintenance costs, and reducing the occurrence of transportation accidents. Double-row tapered roller bearings are used as railway freight car bearings, and their early faults are often mixed with noise, making it difficult or inaccurate to extract fault signals, making it impossible to accurately identify the fault type.

[0003] Vibration noise spectrum analysis is one of the most commonly used rolling bearing fault diagnosis and analysis methods, and signal processing is the most important part of vibration noise spectrum analysis. In existing methods, empirical mode decomposition (EMD) is often used to improve the end-point mode aliasing phenomenon, but its end-point effect cannot be solved; variational mode decomposition VMD (Variational Model Decomposition) can improve the end-point effect and mode aliasing phenomenon, but it is easily affected by the number of decomposition layers and penalty factors. At present, the value of the number of decomposition layers can only be determined by observing whether the center frequency is stable, which is easily affected by human interference and the signal processing effect is unstable.

[0004] Therefore, a railway freight car bearing fault diagnosis method is needed that can improve the endpoint effect and modal aliasing phenomenon. Summary of the invention

[0005] In view of this, the present invention provides a railway freight car bearing fault diagnosis method based on EEMD and BWO-VMD to solve the problem that the railway freight car bearing fault characteristic information cannot be accurately extracted under noise interference.

[0006] To this end, the present invention provides the following technical solutions:

[0007] A railway freight car bearing fault diagnosis method based on EEMD and BWO-VMD, comprising:

[0008] Collect the original vibration signal of truck bearings;

[0009] Reconstruct the original signal through EEMD;

[0010] Several IMF components are obtained by decomposing and reconstructing the signal through BWO-VMD;

[0011] The envelope demodulation is performed using the non-aliased IMF component to obtain the envelope spectrum frequency diagram;

[0012] Fault diagnosis is performed by matching the envelope spectrum frequency diagram with the characteristic frequency of the bearing fault.

[0013] Further, reconstructing the original signal through EEMD includes:

[0014] The original signal of truck bearing vibration is decomposed into several IMF components by EEMD algorithm;

[0015] The effective IMF components containing fault impact information are screened out through correlation coefficient and variance contribution rate;

[0016] The signal is reconstructed by summing all valid IMF components.

[0017] Furthermore, the signal is decomposed and reconstructed by BWO-VMD to obtain several IMF components, including:

[0018] Obtain the optimal parameter set [k, α] of VMD through BWO;

[0019] Substituting the value [k, α] of the optimal parameter group into VMD, the adaptive decomposition of the signal reconstructed by EEMD is performed to obtain several modal component IMFs.

[0020] Furthermore, the obtaining of the optimal parameter group [k, α] of VMD through BWO includes:

[0021] The fitness value is calculated with the minimum envelope entropy as the objective function, that is, the envelope entropy value of each beluga whale's current position;

[0022] Update the position of each beluga whale through the balance factor Bf and the whale fall probability Wf;

[0023] The optimal solution is found by comparing its envelope entropy values.

[0024] Furthermore, the position of each beluga whale is updated by the balancing factor Bf and the whale fall probability Wf:

[0025]

[0026] Where t is the current iteration number; T is the maximum iteration number; B0 is the value randomly generated in each iteration, B0∈(0,1);

[0027] The stage each beluga whale enters is determined by the size of Bf. Different stages correspond to different update mechanisms:

[0028] When Bf>0.5, the White Whale update corresponds to the exploration stage:

[0029]

[0030] In the formula, is the new position of the ith beluga whale in the jth dimension; p is a random integer. Assuming that the optimal dimension of the problem to be solved is d, the value range of p is [1, d]; For the i-th beluga whale at p j Position in dimension; and Represent the current positions of the i-th beluga whale and the r-th beluga whale respectively; r is a random integer in [1, N], N is the population size; N and r 2 is a random operator between (0,1); sin(2πr 2 ) and cos(2πr 2 ) is a mirror image of a beluga whale with its fins facing the water surface;

[0031] When Bf≤0.5, the White Whale update corresponds to the development stage:

[0032]

[0033] In the formula, is the optimal position; r 3 and r 4 is a random number ranging from (0,1); C 1 is the random jump intensity, which is used to measure the Levy flight intensity.

[0034] Furthermore, the bearing fault characteristic frequency includes an outer ring fault frequency and a rolling element fault frequency;

[0035] The outer ring fault frequency:

[0036]

[0037] The rolling element failure frequency:

[0038]

[0039] Among them, fBFI is the characteristic frequency of inner ring fault, fBFB is the characteristic frequency of rolling element fault, r is the bearing speed, n is the number of bearing balls, d is the rolling element diameter, D is the bearing pitch diameter, and α is the rolling element contact angle.

[0040] Advantages and positive effects of the present invention:

[0041] The optimal parameter group of the existing VMD decomposition is usually selected based on manual experience, which is not accurate enough. The present invention uses the White Whale optimization algorithm to optimize the parameters of VMD, which can not only remove the influence of human factors, but also improve the accuracy of the entire diagnosis process. In addition, the BWO iteration convergence speed is relatively faster, the time used for one iteration is shorter, the fault diagnosis efficiency is higher, the fault feature information is more obvious, and the fault diagnosis accuracy is also higher. The EEMD preprocessing method is used to reduce the influence of noise interference, thereby improving the stability of the optimization process and the accuracy of the diagnosis process. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0043] Figure 1 This is a flow chart of a method according to an embodiment of the present invention;

[0044] Figure 2 This is a flow chart of EEMD in an embodiment of the present invention;

[0045] Figure 3 This is a BOW-VMD flow chart in an embodiment of the present invention;

[0046] Figure 4 This is a flow chart of BOW outputting an optimal parameter group in an embodiment of the present invention;

[0047] Figure 5 FIG. 4 is a flow chart of VMD decomposition signals in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0049] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0050] Although vibration signal analysis based on time domain and frequency domain has been widely used in fault diagnosis, analysis from the time domain can only diagnose the approximate location of the bearing fault, and analysis from the frequency domain will lose the frequency domain characteristics. Therefore, analysis in the time and frequency domains can make up for the defects and shortcomings of the first two methods.

[0051] The present invention provides a railway freight car bearing fault diagnosis method based on EEMD and BWO-VMD, and the bearing fault diagnosis is performed by integrating empirical mode decomposition and optimizing variational mode decomposition by white whale algorithm. EEMD can decompose the signal into a finite number of modal functions (IMF) according to the time scale characteristics of the signal itself. The bearing vibration signal with nonlinear and non-stationary characteristics becomes stable, has the advantages of intuitiveness and strong adaptability, and can solve the problem of modal aliasing generated when compared with EMD decomposition. In order to separate the fault characteristics buried by noise, VMD with strong nonlinear, non-stationary and anti-noise performance is selected as the main analysis method for railway freight car bearing vibration signals. And the intelligent optimization algorithm BWO is introduced to obtain the decomposition layer and penalty factor. BWO is a population-based meta-heuristic algorithm proposed according to the white whale behavior, which has an adaptive balance factor and whale fall probability, and plays an important role in controlling exploration and development capabilities. And the Levy flight strategy is introduced in the development stage, which further enhances the global convergence of this stage.

[0052] Combination Figure 1 Flowchart further illustrates the steps of the method of the present invention:

[0053] S1. Use the railway freight car bearing dynamic performance tester to collect the vibration signal of the freight car bearing;

[0054] S2. Preprocessing the collected vibration signal to achieve noise reduction of the vibration signal;

[0055] Combination Figure 2EEMD flow chart, further explanation of vibration signal preprocessing:

[0056] S21, using EEMD to decompose the bearing vibration signal into several IMF components;

[0057] When the vibration signal x(t) is decomposed by EEMD, Gaussian white noise of different amplitudes is added to the vibration signal multiple times to change the signal extreme point characteristics, making its distribution more uniform, thereby improving the modal aliasing phenomenon. Then, the IMF components obtained after multiple EMD decompositions are averaged overall, thereby eliminating most of the noise influence and obtaining accurate IMF components. The specific decomposition process is as follows:

[0058] Step 1: Randomly add Gaussian white noise of different amplitudes to the bearing vibration signal x(t):

[0059] xi(t)=x(t)+kni(t)

[0060] Where k is the added noise amplitude coefficient.

[0061] Step 2: Decompose the new signal using EMD to obtain n IMF components cij(t) and remainder ri(t);

[0062]

[0063] Step 3: Repeat steps 1 and 2 N times, adding Gaussian white noise of different amplitudes each time;

[0064] Step 4 takes the average value of each component as the result, which can remove the influence of Gaussian white noise and obtain accurate IMF components:

[0065]

[0066] Where N is the number of times noise is added; cij(t) is the jth IMF component obtained by EEMD decomposition of the original signal x(t).

[0067] S22, using the correlation coefficient and the variance contribution rate as screening indicators, the screening indicators select the effective IMF components containing the fault impact information and reconstruct the new signal;

[0068] After the original vibration signal of the bearing is preprocessed and denoised by the EEMD algorithm, several modal components IMF are obtained. Some of these components have a high degree of correlation with the original vibration signal of the bearing and contain a lot of effective fault information; while the remaining components have a low degree of correlation with the original vibration signal and contain almost no effective fault information. In order to ensure the quality of the reconstructed signal, it is necessary to screen out the IMF components that are most similar to the original vibration signal and contain the most effective fault information from the obtained IMF components for reconstruction. The remaining invalid components with weak correlation and the noise interference information they contain are removed from the reconstructed signal, thereby achieving the effect of eliminating some noise interference.

[0069] Correlation coefficient: Taking the original signal and the IMF component signal as two variables, the larger the correlation coefficient between the two signals, the stronger the correlation, and the more fault feature information contained in the IMF component signal; conversely, the smaller the correlation coefficient, the weaker the correlation, and the less fault feature information contained. The expression of the correlation coefficient r is as follows:

[0070]

[0071] Where x is the original vibration signal; is the arithmetic mean of x, y is the IMF component signal; is the arithmetic mean of y, m is the number of IMF components of the signal.

[0072] In this embodiment, the IMF components with correlation coefficients greater than 0.3 are selected as preliminary effective components.

[0073] Variance contribution rate: an indicator used to measure the contribution of a variable to the overall variance, which is often used for feature selection or feature importance assessment. A sensor is used to sample physical quantities within a certain period of time at a specific sampling frequency to obtain a discrete signal X(t), where N discrete data are x1, x2…xN. The variance contribution rate of X(t) can be expressed as:

[0074]

[0075] Where μ is the mean of the discrete signal X(t); M is the number of IMF components.

[0076] In this embodiment, the variance contribution rate of the preliminary effective component is calculated, and when the variance contribution rate of the preliminary effective component is greater than the sum of the variance contribution rates of the remaining components, the preliminary effective component is determined as the final effective component.

[0077] The reconstructed signal is obtained by summing all valid components; the reconstructed signal removes the noise interference information in the invalid components. In the time domain diagram of the reconstructed signal, some values ​​have obvious impacts; in the spectrum diagram, the fault characteristic frequency that can be found before reconstruction has an increased amplitude after reconstruction.

[0078] S3, input the reconstructed vibration signal into BWO-VMD, and adaptively decompose the reconstructed new signal;

[0079] S31, Combination Figure 3 The BWO-VMD signal decomposition process includes:

[0080] Taking the minimum envelope entropy as the objective function, BWO is used to optimize the optimal parameter group of VMD. The optimal parameter group is brought into VMD to realize the adaptive decomposition of VMD for reconstructing new signals.

[0081] Combination Figure 4 Further explanation of the optimal parameter group for BWO to optimize VMD:

[0082] Step 1: Initialize BWO parameters, including population size, maximum number of iterations, and value range of target parameters;

[0083] Step 2: The initial position of the beluga population is randomly generated. The adaptive value is calculated through the objective function, that is, the envelope entropy value of the current position of each beluga. The optimal solution is found by comparing the envelope entropy values.

[0084] Step 3: Calculate the balance factor Bf and whale fall probability Wf:

[0085]

[0086] Where t is the current iteration number; T is the maximum iteration number; B0 is the value randomly generated in each iteration, B0∈(0,1);

[0087] Step 4: Continuously update the position of the white whale, local optimal solution and overall optimal solution;

[0088] The stage each beluga whale enters is determined by the size of Bf. Different stages correspond to different update mechanisms:

[0089] When Bf>0.5, the White Whale update corresponds to the exploration stage:

[0090]

[0091] In the formula, is the new position of the ith beluga whale in the jth dimension; p is a random integer. Assuming that the optimal dimension of the problem to be solved is d, the value range of p is [1, d]; For the i-th beluga whale at p jPosition in dimension; and Represent the current positions of the i-th beluga whale and the r-th beluga whale respectively; r is a random integer in [1, N], N is the population size; N and r 2 is a random operator between (0,1); sin(2πr 2 ) and cos(2πr 2 ) is a mirror image of a beluga whale with its fins pointing toward the water.

[0092] When Bf≤0.5, the White Whale update corresponds to the development stage:

[0093]

[0094] In the formula, is the optimal position; r 3 and r 4 is a random number ranging from (0,1); C 1 is the random jump intensity, and the expression used to measure the Levy flight intensity is: C 1 =2r 4 (1-t / T) is a random number that conforms to the Levy distribution, and its expression is as follows:

[0095]

[0096] Where β=1.5, μ and υ are random numbers that obey the normal distribution.

[0097] Calculate the envelope entropy of the beluga whale in the new position and find the optimal solution for the current iteration. The probability of the whale falling needs to be calculated in each iteration.

[0098] Step 5: Determine whether the current number of iterations has reached the maximum number of iterations. If it is greater than or equal to the maximum number of iterations, stop the optimization search and output the optimal result. If it is less than the maximum number of iterations, return to step 3 to continue a new round of optimization search.

[0099] Step 6: Substitute the value [k, α] of the optimal parameter group into VMD to realize the adaptive decomposition of VMD on the reconstructed new signal; k is the number of decomposition layers, and α is the penalty factor; therefore, k modal component IMFs are obtained.

[0100] S32, Combination Figure 5 The VMD flow chart shown in the figure shows that the process of decomposing the signal is as follows:

[0101] Step 1 Initialization {λ 1} and n=0;

[0102] Step 2: Set n=n+1 and start executing the loop;

[0103] Step 3: Let k = 1, according to the formula:

[0104]

[0105] In the formula, Corresponding to μ i (t), x(t), λ(t), and update the parameters

[0106] Step 4: Set k = k + 1, and repeat step 3 until k = K.

[0107] Wherein, K is a hyperparameter, and in this embodiment, the approximate reference range is [2,10];

[0108] Step 5 For all ω ≥ 0, according to the formula: Update the Lagrange multiplier parameter λ;

[0109] In the formula, τ is the fidelity coefficient; ω is the frequency; ∧ is the Fourier transform; n is the number of iterations, and the parameter λ is updated;

[0110] Step 6 Repeat steps 2 to 5 until the convergence condition is met, that is, If the conditions are met, the process ends; otherwise, the process returns to step 2.

[0111] S4. Select the non-aliased IMF component for envelope demodulation to extract the fault feature information of the bearing.

[0112] S5. Obtain a judgment result based on matching the selected envelope spectrum frequency diagram with the characteristic frequency of the bearing fault.

[0113] Among them, the method for obtaining the characteristic frequency of bearing fault is as follows:

[0114] The basic bearing information of the target truck is collected, including: the number of rolling elements, bearing contact angle, bearing inner diameter, bearing outer diameter, outer ring height and bearing speed information; and the speed information is used to calculate the fault frequency, including: outer ring fault frequency and rolling element fault frequency.

[0115]

[0116]

[0117] Among them, fBFI is the characteristic frequency of inner ring fault, fBFB is the characteristic frequency of rolling element fault, r is the bearing speed, n is the number of bearing balls, d is the rolling element diameter, D is the bearing pitch diameter, and α is the rolling element contact angle.

[0118] The present invention introduces the BWO algorithm to obtain the optimal values ​​of the decomposition layer number and the penalty factor, and realizes the adaptive decomposition of the VMD algorithm. Compared with the traditional algorithm, BWO has the advantages of high-performance global search capability, better stability and convergence accuracy.

[0119] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A railway freight car bearing fault diagnosis method based on EEMD and BWO-VMD, characterized in that: include: Collect the original vibration signal of truck bearings; Reconstruct the original signal through EEMD; Several IMF components are obtained by decomposing and reconstructing the signal through BWO-VMD; The envelope demodulation is performed using the non-aliased IMF component to obtain the envelope spectrum frequency diagram; Fault diagnosis is performed by matching the envelope spectrum frequency diagram with the characteristic frequency of the bearing fault.

2. The railway freight car bearing fault diagnosis method based on EEMD and BWO-VMD according to claim 1, characterized in that: The reconstructing the original signal by EEMD comprises: The original signal of truck bearing vibration is decomposed into several IMF components by EEMD algorithm; The effective IMF components containing fault impact information are screened out through correlation coefficient and variance contribution rate; The signal is reconstructed by summing all valid IMF components.

3. The railway freight car bearing fault diagnosis method based on EEMD and BWO-VMD according to claim 1 is characterized in that: The signal is decomposed and reconstructed by BWO-VMD to obtain several IMF components, including: Obtain the optimal parameter set [k, α] of VMD through BWO; Substituting the value [k, α] of the optimal parameter group into VMD, the adaptive decomposition of the signal reconstructed by EEMD is performed to obtain several modal component IMFs.

4. The railway freight car bearing fault diagnosis method based on EEMD and BWO-VMD according to claim 3 is characterized in that: The method of obtaining the optimal parameter group [k, α] of VMD through BWO includes: The fitness value is calculated with the minimum envelope entropy as the objective function, that is, the envelope entropy value of each beluga whale's current position; Update the position of each beluga whale by balancing the factor Bf and the whale fall probability Wf; The optimal solution is found by comparing its envelope entropy values.

5. The railway freight car bearing fault diagnosis method based on EEMD and BWO-VMD according to claim 4 is characterized in that: The position of each beluga whale is updated by balancing the factor Bf and the whale fall probability Wf: Where t is the current iteration number; T is the maximum iteration number; B0 is the value randomly generated in each iteration, B0∈(0,1); The stage each beluga whale enters is determined by the size of Bf. Different stages correspond to different update mechanisms: When Bf>0.5, the White Whale update corresponds to the exploration stage: In the formula, is the new position of the ith beluga whale in the jth dimension; p is a random integer. Assuming that the optimal dimension of the problem to be solved is d, the value range of p is [1, d]; For the i-th beluga whale at p j Position in dimension; and Represent the current positions of the i-th beluga whale and the r-th beluga whale respectively; r is a random integer in [1, N], N is the population size; N and r2 are random operators between (0, 1); sin(2πr2) and cos(2πr2) are the mirror image beluga whales with their fins facing the water surface; When Bf≤0.5, the White Whale update corresponds to the development stage: In the formula, is the optimal position; r3 and r4 are random numbers ranging from (0,1); C1 is the random jump intensity, which is used to measure the Levy flight intensity.

6. The railway freight car bearing fault diagnosis method based on EEMD and BWO-VMD according to claim 1, characterized in that: The bearing fault characteristic frequency includes outer ring fault frequency and rolling element fault frequency; The outer ring fault frequency: The rolling element failure frequency: Among them, fBFI is the characteristic frequency of inner ring fault, fBFB is the characteristic frequency of rolling element fault, r is the bearing speed, n is the number of bearing balls, d is the rolling element diameter, D is the bearing pitch diameter, and α is the rolling element contact angle.

Images