A vibration fault diagnosis method and system based on Q-factor wavelet transform

Abstract

The present application discloses a vibration fault diagnosis method and system based on Q-factor wavelet transform. The rotation machinery vibration fault diagnosis method based on adjustable Q-factor wavelet analysis includes the following steps: obtaining vibration data of the rotation machinery; performing adjustable Q-factor wavelet transform on the vibration data, establishing an AR model of the vibration data, and calculating the fractal features of the vibration data; calculating a feature vector of the data to be measured according to the transformed signal; using a linear discriminant analyzer to classify the feature vectors of the normal and the data to be measured; and judging whether the vibration of the mechanical equipment is normal according to the classification result. Compared with the existing fault analysis methods, the present application has the ability to process non-linear and non-stationary vibration signals, and achieves the technical effect of accurately judging and positioning mechanical vibration faults.

Description

Technical Field

[0001] This application relates to the field of digital processing technology, and particularly relates to a vibration fault diagnosis method and system based on Q-factor wavelet transform. Background Art

[0002] The vibration signal of mechanical equipment can reflect its operating state. At present, the mechanical equipment fault diagnosis methods based on vibration signal analysis mainly include time-domain analysis method and frequency-domain analysis method. The Q-factor wavelet analysis combines the advantages of time-domain analysis and frequency-domain analysis. It is a signal processing technology that integrates multiple information and is a new fault diagnosis and analysis method.

[0003] When a mechanical equipment fails, especially faults such as gear wear, pitting, and tooth breakage, its vibration signal shows a non-linear and non-stationary signal. Although traditional signal processing methods can effectively analyze the signal in the time domain and frequency domain respectively, they lack the ability to analyze the time domain and frequency domain simultaneously. Wavelet transform can combine the advantages of time-domain and frequency-domain analysis, and can be analyzed in both the time domain and frequency domain simultaneously. For time-varying signals, effective inverse transforms in the time domain and frequency domain can be performed. Although wavelet transform combines the characteristics of the time domain compared with Fourier transform, in the transformation process, an effective selection of the wavelet basis is required, and the wavelet function derived from a single basis function cannot accurately reflect the local characteristics of the time-varying signal, and the reconstructed time-domain signal will lose the time-domain characteristics of the original signal. The Tunable Q-Factor Wavelet Transform (TQWT) contains adjustable parameters of the oscillation characteristics of the signal and is a kind of discrete wavelet transform method, which is very suitable for the time-varying signals of mechanical vibration faults. TQWT can effectively represent the non-stationarity of the signal on the time scale. Summary of the Invention

[0004] This application provides a vibration fault diagnosis method based on Q-factor wavelet transform, including the following steps:

[0005] Obtain the vibration data of the rotating machinery;

[0006] Perform tunable Q-factor wavelet transform on the vibration data and establish an AR model of the vibration data;

[0007] Calculate the feature vector of the data to be measured according to the transformed signal;

[0008] Use a linear discriminant analyzer to classify the feature vectors of normal and measured data;

[0009] Judge whether the vibration of the mechanical equipment is normal according to the classification result.

[0010] A vibration fault diagnosis method based on Q-factor wavelet transform as described above, wherein wavelet energy coefficients are constructed using vibration data and used as one of the elements of the data feature vector.

[0011] A vibration fault diagnosis method based on Q-factor wavelet transform as described above, wherein an n-order AR model is established based on the data, and the AR model coefficients are calculated according to the direct estimation method, matrix recursive estimation method, and parameter recursive estimation method.

[0012] A vibration fault diagnosis method based on Q-factor wavelet transform as described above, wherein the fractal dimension of the data is used as one of the elements of the data feature vector.

[0013] A vibration fault diagnosis method based on Q-factor wavelet transform as described above, wherein different combinations of the adjustable Q-factor wavelet coefficients, AR model parameters, and fractal dimensions of the data are used as the data feature vector.

[0014] A vibration fault diagnosis method based on Q-factor wavelet transform as described above, wherein the feature vector is classified using a linear classifier, and the optimal classification threshold, i.e., the mutual information threshold, is calculated, and whether the vibration of the mechanical equipment is normal is judged according to the mutual information threshold.

[0015] A vibration fault diagnosis method based on Q-factor wavelet transform as described above, wherein judging whether the vibration of the mechanical equipment is normal further includes:

[0016] The feature vector is classified using a linear classifier, and the optimal classification threshold is calculated;

[0017] According to the normal data and the data to be measured, the Euclidean distance between the feature vectors of the two sets of data is calculated;

[0018] Whether the mechanical equipment has a fault is judged according to the relative change between the optimal classification threshold and the Euclidean distance.

[0019] The present invention also provides a vibration fault diagnosis system based on Q-factor wavelet transform, including: a rotating machine, a sensor, and a processor;

[0020] Wherein, the sensor: is used to collect the vibration data of the rotating machine and upload the collected vibration data to the processor for processing; the vibration data includes vibration signals from three directions of the rotating machine;

[0021] The processor: is used to receive the vibration data and execute a vibration fault diagnosis method based on Q-factor wavelet transform as described in any one of the above, process the data, and obtain a diagnosis result.

[0022] A vibration fault diagnosis method and system based on Q-factor wavelet transform of the present application can process weak or non-stationary fault feature signals generated by vibrating devices, and can be applicable to different fields such as different sampling times, sample sizes, and sampling frequencies, give the fault features of the devices, and achieve accurate judgment of mechanical faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present application. For those of ordinary skill in the art, other drawings can also be obtained according to these drawings.

[0024] Figure 1 and Figure 2 is a flowchart of a rotating machinery fault diagnosis method based on Q-factor wavelet transform provided by an embodiment of the present invention;

[0025] Figure 3 is a schematic diagram of the TQWT method passing through high-pass and low-pass filters layer by layer;

[0026] Figure 4 is a sample feature classification diagram. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0028] The purpose of the present application is to provide a rotating machinery vibration fault diagnosis method and system based on Q-factor wavelet transform, which can process weak or non-stationary fault feature signals generated by vibrating devices, and can be applicable to different fields such as different sampling times, sample sizes, and sampling frequencies, give the fault features of the devices, and achieve accurate judgment of mechanical faults.

[0029] The present application provides a rotating machinery fault diagnosis system based on Q-factor wavelet transform, including: a rotating machinery, a sensor, and a processor;

[0030] Among them, the sensor: is used to collect the vibration data of the rotating machinery and upload the collected vibration data to the processor for processing; the vibration data includes vibration signals in three directions from the rotating machinery.

[0031] Processor: It is used to receive the vibration data uploaded by the sensor, and execute the following rotating machinery vibration fault diagnosis method based on Q-factor wavelet transform to process the vibration data and obtain the diagnosis result.

[0032] To achieve the above object, as Figure 1 、 2 shown, this application provides a rotating machinery fault diagnosis method based on Q-factor wavelet transform, including the following steps:

[0033] S1: Obtain the vibration data of the rotating machinery.

[0034] Specifically, data acquisition is performed on the rotating machinery to be fault diagnosed through a sensor to obtain vibration data, and the data is uploaded to the processor. Among them, the vibration data includes vibration signals in three directions from the rotating machinery.

[0035] S2: Use TQWT (tunable Q-factor wavelet transform) to transform the vibration data and establish its AR model.

[0036] S21: Establish high-pass and low-pass filters;

[0037] According to the quality factor Q and the oversampling rate r, that is, determine the parameters α i , β i of the high-pass and low-pass filters according to the equation, and establish the high-pass and low-pass filters according to the equation.

[0038] S22: Perform layer-by-layer high-pass and low-pass filtering on the original signal.

[0039] Specifically, use but not limited to the rectangular window function to divide the multi-dimensional time series to obtain the data set of the original vibration signal. Divide the vibration signal layer by layer through the high-pass and low-pass filters according to the Figure 3 shown TQWT method. Figure 3 where S(n) is the original vibration signal. After the signal S(n) is subjected to discrete Fourier transform, discrete wavelet transform is performed on the vibration data through two-channel filtering and scale transformation, and finally decomposed into a high-pass signal d i (n) and a low-pass signal c i (n). The high-pass signal is used as the output signal of this decomposition layer, and the low-pass signal is used as the input signal of the next layer of decomposition for further decomposition. Decompose the signal until the specified number of layers J is reached, and finally obtain J + 1 wavelet transform coefficients of the vibration signal at each window.

[0040] In the tunable Q-factor wavelet transform of the i-th layer, the sampling frequencies of the high-pass signal d i (n) and the low-pass signal c i (n) are both less than the sampling frequency of the original signal, which are α i f s and βi f s , where 0 < α i , β i < 1.

[0041] Specifically, in the tunable Q-factor wavelet transform of the i-th layer, the expressions of the high-pass and low-pass filters are as follows:

[0042]

[0043]

[0044] Among them, α i is the parameter of the high-pass filter, β i is the parameter of the low-pass filter, ω is the frequency, and the θ function represents the Daubechies frequency response, and its expression is as follows:

[0045]

[0046] Among them, the quality factor Q and the oversampling rate r of the tunable Q-factor wavelet transform can be represented by the parameters α i , β i as follows:

[0047]

[0048] Repeat sub-step S22 to obtain the wavelet transform coefficients of all vibration signal segments in all directions.

[0049] S23: Establish an AR model (autoregressive model) of the vibration signal;

[0050] Specifically, determine the order and AR coefficients of the AR model of each vibration signal segment through cross-validation.

[0051] Repeat sub-step S23 to obtain the AR model coefficients of the vibration signal segments in all directions.

[0052] S3: Calculate the fractal number feature of each vibration signal segment.

[0053] S4: Establish the feature vectors of the normal data and the data to be measured according to the Q-factor wavelet transform coefficients, AR model coefficients, and fractal number features.

[0054] Specifically, nonlinear features in the time-frequency domain, such as wavelet coefficient energy, AR coefficient and fractal dimension, are calculated according to the transformed signal, which specifically includes: extracting features from the decomposed sub-bands using a sliding rectangular window, wherein the features are divided into three categories: wavelet coefficient energy, AR coefficient and fractal dimension, the number of wavelet coefficient energies is J+1, the AR order is n, there are n AR coefficients, and 2 fractal dimensions, so the number of feature quantities is: J+n+3.

[0055] Among them, the calculation methods of the three types of parameters are as follows:

[0056] The calculation formula of the wavelet coefficient energy of the signal is: K is the number of data in the data segment. The n-order AR estimation model of the signal is e(t) is the residual, and the AR coefficient a(i) in the equation is used as the element of the eigenvector. Fractal dimension is a statistical definition of the complexity and nonlinear characteristics of time domain signals, which can explain the irregularity and instability of signals. The box dimension calculation method is used to calculate the fractal dimension, and the maximum side length of the grid is selected as 1024, or the maximum side length can be 2 n ,n≥10.

[0057] S5: Use a linear discriminant classifier to classify the normal and test data feature vectors;

[0058] Specifically, a linear discriminant classifier is used to classify the normal and test data feature vectors to determine the optimal classification threshold, that is, the mutual information threshold. Among them, the classification using the linear discriminant analysis method includes:

[0059] 1) Define the Fisher criterion function:

[0060]

[0061] Among them, Ω represents the direction vector of the projection, S b and S Ω Represents the inter-class and intra-class dispersion matrices of two categories: normal vibration data samples and test data samples.

[0062] 2) Use the Lagrange multiplier method to solve the extreme value of the function;

[0063] L(Ω,λ)=Ω T S b Ω-λ(Ω T S Ω Ω-C) C is an arbitrary constant (6)

[0064] 3) Taking partial derivative of the equation with respect to Ω, the optimal projection direction is:

[0065]

[0066] wherein and are the mean vectors of the normal data samples and the data samples to be measured, respectively.

[0067] S6: Determine whether the vibration of the mechanical equipment is normal according to the mutual information threshold;

[0068] The sub-steps of determining whether the vibration of the mechanical equipment is normal according to the mutual information threshold include: calculating the magnitude of the mutual information according to the mutual information calculation formula:

[0069] MI = 0.5log2(SNR t ) + 1 (8)

[0070] where SNR t represents the signal-to-noise ratio, and its calculation formula is:

[0071]

[0072] where varN represents the variance of the normal vibration data set, varD represents the variance of the data set to be detected, and varA represents the variance of all data sets. The larger the mutual information value, the better the classification effect of the classification system. When the mutual information value is greater than the mutual information threshold, it can be determined that there is an obvious classification between the data to be detected and the normal data, and it can be determined that the vibration of the mechanical equipment is abnormal.

[0073] In addition, the present application can also determine whether the mechanical equipment has a fault through the following scheme, specifically including: calculating the Euclidean distance between the feature vector of the normal data and the feature vector of the data to be measured; determining whether the mechanical equipment has a fault through the relative transformation of the optimal classification threshold and the Euclidean distance.

[0074] The present application has the technical effects of being able to process weak or non-stationary fault feature signals generated by vibration equipment, being applicable to different fields such as different sampling times, sample sizes, and sampling frequencies, giving the fault features of the equipment, and realizing the accurate judgment of mechanical faults.

[0075] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications once they know the basic creative concepts. Therefore, the protection scope of the present application is intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application. Obviously, those skilled in the art can make various changes and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the protection of the present application and its equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. A vibration fault diagnosis method based on Q-factor wavelet transform, characterized in that It includes the following steps: Obtain the vibration data of the rotating machinery; the vibration data includes vibration signals from three directions of the rotating machinery; Perform an adjustable Q-factor wavelet transform on the vibration data, establish an AR model of the vibration data, and calculate the fractal characteristics of the vibration data; Establish the feature vectors of the normal data and the data to be measured according to the Q-factor wavelet transform coefficients, AR model coefficients, and fractal dimension characteristics; Use a linear discriminant classifier to classify the feature vectors of the normal data and the data to be measured; Judge whether the vibration of the mechanical equipment is normal according to the classification result; Among them, use a linear classifier to classify the feature vectors of the normal data and the data to be measured, calculate the mutual information threshold, and judge whether the vibration of the mechanical equipment is normal according to the mutual information threshold; Among them, the sub-steps of judging whether the vibration of the mechanical equipment is normal according to the mutual information threshold include: calculate the mutual information value according to the mutual information calculation formula: ； Among them, represents the signal-to-noise ratio, and its calculation formula is: ； Among them, represents the variance of the normal vibration data set, represents the variance of the data set to be detected, represents the variance of all data sets; When the mutual information value is greater than the mutual information threshold, it is judged that the vibration of the mechanical equipment is abnormal.

2. The vibration fault diagnosis method based on Q-factor wavelet transform according to claim 1, wherein, Use the vibration data to construct Q-factor wavelet transform coefficients as one of the elements of the data feature vector.

3. A vibration fault diagnosis method based on Q-factor wavelet transform according to claim 1, characterized in that, Establish its order AR model based on vibration data, and calculate the AR model coefficients according to the direct estimation method, matrix recursive estimation method, and parameter recursive estimation method.

4. A vibration fault diagnosis method based on Q-factor wavelet transform according to claim 1, characterized in that Use the fractal dimension of the vibration data as one of the elements of the data feature vector.

5. A vibration fault diagnosis method based on Q-factor wavelet transform according to claim 1, characterized in that Use different combinations of the Q-factor wavelet transform coefficients, AR model coefficients, and fractal dimension characteristics of the vibration data as the feature vectors of the normal data and the data to be measured.

6. The vibration fault diagnosis method based on Q-factor wavelet transform according to claim 1, wherein Judging whether the vibration of the mechanical equipment is normal also includes: Use a linear classifier to classify the feature vectors and calculate the optimal classification threshold; Calculate the Euclidean distance of the feature vectors of the two sets of data according to the normal data and the data to be measured; Judge whether the mechanical equipment has a fault according to the relative change between the optimal classification threshold and the Euclidean distance.

7. A vibration fault diagnosis system based on Q-factor wavelet transform, characterized in that, It includes: Rotating machinery, sensors, and processors; Among them, the sensor: is used to collect the vibration data of the rotating machinery and upload the collected vibration data to the processor for processing; the vibration data includes vibration signals from three directions of the rotating machinery; The processor: is used to receive the vibration data and execute a vibration fault diagnosis method according to any one of claims 1-6, process the vibration data, and obtain a diagnosis result.

Images