Method for fault diagnosis of multi-channel sliding bearing acoustic emission signal based on multivariate variational modal decomposition and full vector envelope analysis

Abstract

This invention proposes a fault diagnosis method for multi-channel acoustic emission signals of sliding bearings based on multivariate variational mode decomposition and full vector envelope analysis. The method includes: installing acoustic emission signal sensors in two directions along the sliding bearing cross-section to collect multi-channel acoustic emission signals under normal operation and different bearing damage states; classifying the multi-channel acoustic emission signals to obtain datasets of multi-channel acoustic emission signals under different operating conditions; performing multivariate variational mode decomposition on the multi-channel acoustic emission signals to obtain multiple intrinsic mode function components in different directions; using correlation kurtosis as a parameter, selecting intrinsic mode function components in each direction with a correlation kurtosis greater than that of the original acoustic emission signal, and recombining them into new acoustic emission signals; performing full vector envelope analysis on the recombined acoustic emission signals; organically fusing the recombined acoustic emission signals in the two directions according to elliptical trajectories; and performing envelope analysis on the fused signal to obtain full vector envelope spectra under normal and different damage states, forming a database for judging sliding bearing bearing damage. The proposed method applies the same processing to the acoustic emission signal of the sliding bearing to be diagnosed, and compares the resulting full-vector envelope spectrum with images in the database to diagnose the damage state of the sliding bearing bush. This invention effectively solves the problems of low accuracy and incomplete results in unidirectional sliding bearing bush fault diagnosis.

Description

Technical Field

[0001] This invention belongs to the field of mechanical system fault diagnosis and signal processing technology. Specifically, it relates to a fault diagnosis method for acoustic emission signals of multi-channel sliding bearings based on multivariate variational mode decomposition and full vector envelope analysis. Background Technology

[0002] In the context of modern large-scale production, sliding bearings are an indispensable component in many mechanical equipment. As key support and lubrication parts in heavy-duty machines such as large wind turbines and hydroelectric generators, they play a crucial role. With increasing service life, internal components of sliding bearings, such as the bearing shell and internal support rings, can suffer damage. Therefore, condition monitoring and fault diagnosis of sliding bearings are of paramount importance. Summary of the Invention

[0003] This invention proposes a fault diagnosis method for multi-channel acoustic emission signals of sliding bearings based on multivariate variational mode decomposition and full vector envelope analysis. The method employs an acoustic emission signal sensor to collect multi-channel acoustic emission signals from two directions at the same cross-section of the sliding bearing. Multivariate variational mode decomposition is performed on the multi-channel acoustic emission signals, and the intrinsic mode function components (IMFs) obtained from the decomposition in the two directions are recombined according to the magnitude of their correlation kurtosis. Suitable IMFs are obtained by using correlation kurtosis as a screening criterion and then recombined, resulting in a relatively pure acoustic emission signal with noise removed. Next, the recombined acoustic emission signals from the two directions are organically fused using a full vector envelope spectrum to obtain a database for assessing sliding bearing bearing damage. The proposed method is applied to the acoustic emission signals of the sliding bearing to be assessed, and the processed full vector envelope spectrum is compared with the images in the database to diagnose the damage state of the sliding bearing bearing. Accurate identification of sliding bearing bearing damage is achieved through the analysis of the full vector envelope spectrum.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0005] (a) Acoustic emission signal sensors are installed in two directions on the cross-section of the sliding bearing. The two acoustic emission signal sensors are located in different directions on the same cross-section, with a 90-degree interval between them. The acoustic emission sensors are connected to a data acquisition card, which transmits data to a computer via a USB interface. The sampling frequency is 1000 kHz, and the continuous sampling time is 100 seconds. Acoustic emission signals of the sliding bearing under normal operation and different bearing damage conditions with different degrees of failure are collected. The obtained acoustic emission signals are classified to obtain an initial dataset including acoustic emission signals under normal operation and different bearing damage conditions.

[0006] (b) The acoustic emission signals under normal conditions and different bearing damage conditions are divided into signal segments with 1e7 signal points per segment. The acoustic emission signals are subjected to multivariate variational mode decomposition to obtain multiple intrinsic mode function components in each direction. The intrinsic mode function components in each direction with a correlation kurtosis as a parameter index are selected and recombined into new acoustic emission signals. The parameters of mode number and penalty factor in multivariate variational mode decomposition are 2 and 1200, respectively.

[0007] (c) Perform full vector envelope analysis on the new acoustic emission signal, and organically fuse the acoustic emission signals in two directions through elliptical trajectory. That is, first perform Hilbert transform on the acoustic emission signals in two directions, then fuse the obtained envelope signals, and finally obtain the final full vector envelope spectrum based on the principal vibration vector of the fused signal. Obtain full vector envelope spectrum images under normal state and different bearing damage states, and then obtain a database for sliding bearing bearing damage diagnosis.

[0008] (d) The acoustic emission signal of the sliding bearing to be diagnosed is processed in the same way, and the processed full vector envelope spectrum is compared with the image in the database to complete the judgment of the damage state of the sliding bearing bush.

[0009] As a preferred technical solution, in step (a), before collecting acoustic emission signal data for each fault state, acoustic emission signal data under normal bearing operation is collected first, and then bearing damage data is collected sequentially according to three bearing damage states: damage, 2mm scratch, and 4mm scratch; after collecting normal data and three types of fault data, the data is preprocessed, including: preliminary noise reduction processing of acoustic emission signals; acoustic emission signals are retained to five decimal places;

[0010] As a preferred technical solution, in step (b), the multi-channel acoustic emission signal is subjected to multivariate variational mode decomposition, the correlation kurtosis of each component in the intrinsic mode function component obtained by decomposition of each channel is calculated and the average correlation kurtosis is obtained. The intrinsic mode function components in each channel with a correlation kurtosis higher than the average correlation kurtosis are screened and recombined into a new multi-channel acoustic emission signal. Compared with the original acoustic emission signal, the periodic impact of this acoustic emission signal is more obvious and the effective information component is higher.

[0011] As a preferred technical solution, in step (c), the recombined multi-channel acoustic emission signal is subjected to full vector envelope analysis. First, the signal is subjected to Hilbert transform to obtain the envelope signals of the two channels. The envelope signals of the two channels are combined into a complex signal, and the sequence of the complex signal is decomposed into the principal vector, the secondary vector and the phase according to the elliptical motion trajectory. The principal vector is expanded along the frequency axis to obtain the full vector envelope spectrum.

[0012] As a preferred technical solution, in step (c), the total vector envelope spectra of the normal state and different bearing damage states are saved sequentially to the same folder according to different states to form a database of different damage states of sliding bearing bearings.

[0013] As a preferred technical solution, in step (d), the same multivariate variational mode decomposition and full vector envelope analysis methods are used to acquire and analyze signals of the sliding bearing to be diagnosed. The obtained full vector envelope spectrum is compared with the images in the database, thereby making a relatively accurate judgment on the bearing condition of the sliding bearing to be diagnosed.

[0014] By adopting the above technical solution, the beneficial effects of the present invention are as follows:

[0015] This invention employs a multi-channel signal decomposition algorithm based on multivariate variational mode decomposition, which can simultaneously decompose signals from two channels into multiple intrinsic mode function components in two directions under the same constraints. While ensuring decomposition accuracy, it decomposes non-stationary and nonlinear signals into relatively stationary sinusoidal signals.

[0016] This invention proposes using correlation kurtosis as an evaluation index for signal screening. By calculating the correlation kurtosis and average correlation kurtosis of multiple intrinsic mode functions in two directions, multiple intrinsic mode function components with higher correlation kurtosis are screened and recombined. This can enhance the impulsive characteristics of the signal while reducing residual noise and increasing the proportion of effective components.

[0017] This invention employs total vector envelope analysis as a fusion analysis algorithm for multi-channel signals. It decomposes the envelope signals obtained after Hilbert transform of acoustic emission signals in two directions according to the elliptical motion trajectory, thereby organically merging the signals from the two channels into one signal. At the same time, the Hilbert transform further improves the stability of the signal, thereby improving the accuracy of the signal while ensuring relatively balanced and comprehensive calculation results. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the embodiments described herein are only for illustration and explanation of the invention and are not intended to limit the present invention.

[0019] Figure 1 This is a flowchart of a fault diagnosis method constructed according to a preferred embodiment of the present invention;

[0020] Figure 2 This is a time-domain image of a slightly damaged sliding bearing bush, collected according to a preferred embodiment of the present invention.

[0021] Figure 3This is a three-dimensional time-domain plot of the intrinsic mode functions after multivariate mode decomposition according to a preferred embodiment of the present invention.

[0022] Figure 4 It is a total vector envelope spectrum obtained after total vector envelope analysis. Detailed Implementation

[0023] 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.

[0024] The terminology used in this embodiment is explained, including: Multivariate Variational Mode Decomposition (MVMD) and Intrinsic Mode Function (IMF).

[0025] like Figure 1 The diagram shows a flowchart of a multi-channel sliding bearing acoustic emission signal fault diagnosis method based on multivariate variational mode decomposition and full vector envelope analysis, constructed according to a preferred embodiment of the present invention. It includes the following steps:

[0026] Step 1: Acquire multi-channel acoustic emission signals of the sliding bearing under normal operation and different bearing damage conditions at a sampling frequency of 1000kHz. Preprocess each acoustic emission signal, including noise reduction and rounding to five decimal places. The time-domain plot of the processed inner ring fault acoustic emission signal is shown below. Figure 2 As shown.

[0027] Step 2: Perform multivariate variational mode decomposition on the acoustic emission signal to obtain several eigenmode function plots in two directions. The eigenmode functions are as follows: Figure 3 As shown.

[0028] Step 3: Calculate the correlation kurtosis value and average correlation kurtosis value of each intrinsic mode function in both directions. Filter and reassemble the intrinsic mode function components with correlation kurtosis values ​​higher than the average correlation kurtosis value to obtain the reassembled acoustic emission signals in both directions.

[0029] Step 4: Perform full vector envelope analysis on the reconstructed acoustic emission signal. The acoustic emission signals in the two directions are organically fused using an elliptical trajectory. Finally, the final full vector envelope spectrum is obtained based on the principal amplitude vector of the fused signal, as shown below. Figure 4 As shown.

[0030] Step 5: Construct full vector envelope spectrum images of the normal state and different bearing damage states to obtain a database for sliding bearing bearing damage diagnosis.

[0031] Step 6: Collect the multi-channel acoustic emission signal of the current sliding bearing, and use the methods in steps 1, 2, 3, and 4 above to obtain the full vector envelope spectrum of the sliding bearing bush to be diagnosed. By comparing it with the database constructed in step 5, the judgment of the damage of the sliding bearing bush to be diagnosed is completed.

[0032] 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 spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A fault diagnosis method for acoustic emission signals of multi-channel sliding bearings based on multivariate variational mode decomposition and full vector envelope analysis, characterized in that, Includes the following steps: (a) By installing acoustic emission signal sensors in two directions on the cross section of the sliding bearing, acoustic emission signals of the sliding bearing under normal operation and under different bearing damage conditions are collected at different fault levels. The obtained acoustic emission signals are classified to obtain an initial dataset including acoustic emission signals under normal operation and under different bearing damage conditions. (b) Perform multivariate variational mode decomposition on the acoustic emission signal to obtain multiple intrinsic mode function components in each direction. Using correlation kurtosis as a parameter, select the intrinsic mode function components in each direction that have a correlation kurtosis greater than that of the original acoustic emission signal and reassemble them into a new acoustic emission signal. (c) Perform full vector envelope analysis on the new acoustic emission signal, and organically fuse the acoustic emission signals in two directions through elliptical trajectory. That is, first perform Hilbert transform on the acoustic emission signals in two directions, then fuse the obtained envelope signals, and finally obtain the final full vector envelope spectrum based on the principal vibration vector of the fused signal, thus completing the noise reduction and processing of acoustic emission signals under normal conditions and different bearing damage conditions. (d) Acquire and process the acoustic emission signal of the sliding bearing under test according to the above steps. The obtained full vector envelope spectrum, together with the full vector envelope spectra under normal conditions and different damage conditions, form a database for the diagnosis of sliding bearing bush damage. (e) The acoustic emission signal of the sliding bearing to be diagnosed is processed in the same way, and the processed full vector envelope spectrum is compared with the image in the database to complete the judgment of the damage state of the sliding bearing bush.

2. The multi-channel sliding bearing acoustic emission signal fault diagnosis method based on multivariate variational mode decomposition and full vector envelope analysis as described in claim 1, characterized in that, In step (a), after collecting the acoustic emission signals of the sliding bearing under normal operation and different bearing damage conditions, the data is preprocessed, including: preliminary noise reduction processing of the acoustic emission signals; the acoustic emission signals are retained to four decimal places.

3. The multi-channel sliding bearing acoustic emission signal fault diagnosis method based on multivariate variational mode decomposition and full vector envelope analysis as described in claim 1, characterized in that, In step (c), a full vector envelope analysis is performed on the acoustic emission signal; the multi-channel envelope signal obtained after Hilbert transformation is defined as the principal vibration vector and amplitude vector of the vibration according to the major and minor semi-axis of the elliptical trajectory, and the final full vector envelope spectrum is obtained through the principal vibration vector.

4. The multi-channel sliding bearing acoustic emission signal fault diagnosis method based on multivariate variational mode decomposition and full vector envelope analysis as described in claim 1, characterized in that, In step (d), the damage status of the sliding bearing bush is determined by comparing the amplitude and frequency in the total vector envelope spectrum obtained from the main vibration vector, especially in the mid-to-high frequency range.

Images