A method for analyzing oil-depleted bearing signals based on frequency centroid exponentiation and minimum entropy deconvolution.

Abstract

A signal analysis method for oil-deficient bearings based on frequency centroid index and minimum entropy deconvolution is proposed. This method evaluates the lubrication condition of bearings using vibration signals. First, the signal to be evaluated undergoes kurtosis analysis and minimum entropy deconvolution to remove large external interferences and amplify the pulse components. Then, envelope spectrum analysis is used to determine whether the pulse components are random pulse signals caused by oil deficiency faults. Subsequently, bearing oil deficiency characteristic parameters are calculated, and finally, the bearing lubrication condition is judged based on a set threshold. This invention can effectively extract random impacts caused by bearing lubrication faults and identify oil deficiency faults in rolling bearings through a dimensionless oil deficiency characteristic parameter (SLF). It is also robust to changing operating conditions and has significant implications for bearing lubrication condition monitoring in engineering applications.

Description

Technical Field

[0001] This invention belongs to the fields of signal processing and analysis technology and bearing condition monitoring, and specifically relates to a signal analysis method for oil-deficient bearings based on frequency centroid index and minimum entropy deconvolution. Background Technology

[0002] Vibration signal-based detection has become the most widely used method for bearing condition monitoring due to its convenience and high sensitivity to early faults. Vibration signals have been extensively studied in detecting fatigue faults such as bearing cracks and spalling, and a series of signal analysis methods have been proposed, such as spectral kurtosis analysis, cyclostationary analysis, wavelet transform, empirical mode decomposition, and cepstral analysis. These methods are very effective in processing periodic vibration signals generated by defective bearings. In contrast, research on monitoring bearing lubrication status through vibration response is relatively limited. Fault signals caused by insufficient lubrication are high-occurrence random vibration signals that are often undetectable using traditional analysis techniques.

[0003] Lubrication of rolling bearings is classified into three states: full film lubrication, mixed lubrication, and boundary lubrication. For bearings lacking oil, the possible lubrication states are mixed lubrication or boundary lubrication. When the thickness of the oil film is less than the height of the rough surface, direct contact between the bearing rollers and the inner and outer raceways will randomly occur. During this contact process, small impacts from surface irregularities will excite the high-frequency vibration level of the bearing. The main component of this response corresponds to the natural frequency of the rolling bearing. Although the vibration signal amplitude is significantly reduced under fully lubricated conditions, the natural frequency still exists. Therefore, it is inappropriate to simply use the presence of the bearing's natural frequency in its natural frequency band to determine the lubrication state.

[0004] Compared to the periodic impact signal caused by bearing spalling failure, the random impact caused by oil shortage failure is much weaker. In noisy environments, the vibration signal of oil shortage failure is easily masked, thus requiring signal amplification. However, the natural frequency of a rolling bearing is related to its material and shape, making it impossible to determine a fixed frequency band for traditional filtering. Both spalling and oil shortage failure impacts are pulse signals. The idea behind Minimum Entropy Deconvolution (MED) is to design an FIR filter that outputs a filtered signal with maximum kurtosis, thus demonstrating good performance in amplifying pulse signals. This invention proposes using kurtosis values ​​to measure the impulsiveness of the signal and utilizing MED to amplify the fault signal.

[0005] The impacts caused by oil shortage faults occur randomly, so classical envelope analysis used for periodic impact detection cannot obtain further diagnostic information. A reliable method for mining the characteristics of random oil shortage impacts is needed. Kurtosis is one of the most popular indicators of fault severity, but it is very sensitive to large interference pulse signals. Root mean square (RMS) is a common measurement, but even with severe faults, it is affected by rotational speed. The center-of-gravity index (SCI) is the weighted average frequency of the signal power spectrum, related to the most dominant natural frequency in the signal. The occurrence of random oil shortage impacts leads to an increase in high-frequency vibration energy, thus disrupting the spectral balance. Therefore, SCI can be used to monitor the increase in high-frequency components of the signal caused by oil shortage. However, SCI is in units of Hz, which makes it difficult to set a universally applicable lubrication condition discrimination threshold. Furthermore, shaft vibration can shift the signal to the low-frequency region, affecting the SCI results. Therefore, developing a dimensionless oil shortage characteristic parameter robust to changing operating conditions is an urgent problem to be solved. Summary of the Invention

[0006] To overcome the above-mentioned shortcomings, the present invention aims to provide a method for analyzing vibration signals of oil-deficient bearings. This method is based on the frequency centroid index and combined with the minimum entropy deconvolution (MED) method, and proposes a dimensionless oil-deficient characteristic parameter (SLF). This parameter can effectively distinguish between the moderate lubrication working state and the oil-deficient lubrication working state of rolling bearings, and has robustness under the variable operating conditions of bearings.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A method for analyzing oil-deficient bearing signals based on frequency centroid exponentiation and minimum entropy deconvolution includes the following steps:

[0009] Step 1: Place the sensor on the bearing mounting base under test and acquire the vibration / acoustic signal measurement value x(n) of the bearing under working conditions, where n = 0, 1, ..., N-1, N is the data length, and the sampling frequency is denoted as Fs;

[0010] Step 2: Perform a Fast Fourier Transform on the original signal x(n) to obtain the signal spectrum, and calculate the centroid index (SCI) of x(n) according to formulas (1) and (2). Raw ;

[0011]

[0012]

[0013] In the formula, the sampling time interval Δt = 1 / Fs, It is the one-sided power spectrum of the original signal x(n).

[0014] Step 3: Perform minimum entropy deconvolution (MED) on the original signal x(n) to obtain the pulse component y(n) in the signal. Calculate the kurtosis value of y(n) to determine whether there are large interference pulses from external sources. Set a threshold value T. Kurt If the kurtosis value is greater than T Kurt If the signal is contaminated, it should be measured again; otherwise, proceed to the next step of analysis.

[0015] Step 4: Calculate the squared envelope and perform discrete Fourier transform on the pulse component y(n) to obtain the envelope spectrum. Check if there are any bearing fault characteristic frequencies (BCFs) in the envelope spectrum. If they exist, it is determined that the bearing has a spalling or cracking fault; if they do not exist, it is determined that the component is a non-periodic impact, i.e., a random impact, and proceed to the next step.

[0016] Step 5: Perform a Fast Fourier Transform on the pulse component y(n) to obtain the signal spectrum, and calculate the centroid index (SCI) of y(n). MED Step two: The frequency centroid index (SCI) of the original signal is known. Raw The oil exhaust characteristic parameter SLF of the working bearing is calculated according to formula (3);

[0017] SLF = SCI MED / SCI Raw (3)

[0018] Step Six: Set the threshold value T SLF And make a judgment; if the characteristic parameter SLF of the exhausted oil is greater than T SLF If the oil level is low, the working bearing is considered to be in a state of insufficient oil operation; otherwise, the working bearing is in a state of proper lubrication.

[0019] The present invention has the following beneficial effects:

[0020] a) The method of the present invention can adaptively distinguish between bearing spalling / crack faults and bearing lubrication faults through envelope spectrum analysis.

[0021] b) This invention combines kurtosis analysis with minimum entropy deconvolution (MED) method to effectively eliminate the influence of external large impacts and noise on diagnostic results, and enhances the random impact signal components caused by lubrication failure.

[0022] c) The method of this invention combines the frequency centroid index (SCI) and minimum entropy deconvolution (MED) to propose a dimensionless oil loss characteristic parameter (SLF), which can effectively identify oil loss faults in rolling bearings and is robust to changing operating conditions, thus helping to monitor bearing lubrication status in engineering applications. Attached Figure Description

[0023] Figure 1 This is a flowchart of the present invention.

[0024] Figure 2 This is a structural diagram of the experimental platform according to an embodiment of the present invention.

[0025] Figure 3 The time-domain waveforms and kurtosis values ​​of the vibration signals of test bearing 1 (left) and test bearing 2 (right) are shown in the embodiments of the present invention.

[0026] Figure 4 The images show the vibration signal spectrum and frequency centroid index calculation results of test bearing 1 (left) and test bearing 2 (right) in this embodiment of the invention.

[0027] Figure 5 The time-domain waveforms and kurtosis values ​​of the MED filtered signals of test bearing 1 (left) and test bearing 2 (right) are shown in this embodiment of the invention.

[0028] Figure 6 The envelope spectra of the MED filtered signals of test bearing 1 (left) and test bearing 2 (right) are shown in this embodiment of the invention.

[0029] Figure 7 The MED filtered signal spectrum diagrams and frequency centroid index calculation results of test bearing 1 (left) and test bearing 2 (right) in this embodiment of the invention are shown.

[0030] Figure 8 This is a graph showing the SLF-speed curve of the bearing tested in an embodiment of the present invention. Detailed Implementation

[0031] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0032] The specific parameters are as follows: 1) Test bearing 1 is an oil-deficient bearing, with only a minimal amount of grease added to prevent the bearing from seizing during rotation; Test bearing 2 is a normally lubricated bearing, with grease filling approximately 1 / 2 of the bearing raceway; 2) The applied load weight is 15kg; 3) The spindle speed is 900r / min; 4) The test bearing model is SKF6306: a) Roller diameter: 12.3mm, b) Pitch diameter: 52mm, c) Contact angle: 0°, d) Number of rollers: 8; 5) The signal sampling frequency Fs is 100kHz;

[0033] The method of this invention is applied to analyze vibration signals and diagnose bearing lubrication faults, such as... Figure 1 As shown, a method for analyzing the signal of a depleted oil bearing based on the frequency centroid index and minimum entropy deconvolution includes the following steps:

[0034] Step 1: As Figure 2Set up the experimental platform and place the accelerometer on the test bearing mounting base. Obtain the vibration signal measurement values ​​x1(n) of test bearing 1 (oil-deficient bearing) and x2(n) of test bearing 2 (normally lubricated bearing) under the same operating conditions (load 15kg, speed 900r / min), where n = 0, 1, ..., N-1. The data lengths of x1(n) and x2(n) are equal, both being N. Figure 3 As shown, the sampling frequency Fs = 100kHz;

[0035] Step 2: Perform Fast Fourier Transform on the original signals x1(n) and x2(n) of the two test bearings respectively to obtain the signal spectrum, and calculate the frequency centroid index of the signal according to formulas (1) and (2) respectively. Figure 4 As shown, SCI 1-Raw =13.9836kHz, SCI 2-Raw =11.9852kHz;

[0036] Step 3: Perform minimum entropy deconvolution (MED) on the original signals x1(n) and x2(n) respectively to obtain the impulse components y1(n) and y2(n) in the signal, and calculate the kurtosis value of each. Figure 5 As shown, the kurtosis values ​​of the two pulse components are Kurtosis1 = 10.63 and Kurtosis2 = 6.82, and the threshold value T is set. Kurt =50, then Kurtosis1 < T Kurt And Kurtosis2 < T Kurt Therefore, the signal was not contaminated, and we can proceed to the next step of analysis.

[0037] Step 4: Calculate the squared envelope of the pulse components y1(n) and y2(n) and perform a discrete Fourier transform to obtain the envelope spectrum. Check if any bearing fault characteristic frequencies (BCFs) exist in the envelope spectrum. Figure 6 As shown, there is no definite BCF, so the detected signal component is identified as a non-periodic impulse, i.e., a random impulse, and further processing is carried out.

[0038] Step 5: Perform Fast Fourier Transform on the pulse components y1(n) and y2(n) to obtain the signal spectrum, and calculate their frequency centroid indices respectively, such as... Figure 7 As shown, SCI 1-MED =14.691kHz, SCI 2-MED =1.7203kHz, Step 2: The centroid index (SCI) of the original signal is known. 1-Raw With SCI 2-Raw According to formula (3), the oil exhaust characteristic parameters of the two test bearings are calculated respectively: SLF1 = 1.051, SLF2 = 0.144;

[0039] Step Six: Set the threshold value T SLF =0.5 and make a judgment, test the oil drain characteristic parameter SLF1 of bearing 1 > T SKF Therefore, it was determined that the bearing was in a state of oil deficiency, and the oil deficiency characteristic parameter SLF2 < T was tested for bearing 2. SLF Therefore, it was determined that the bearing was in a proper lubrication condition, and the analysis results were consistent with the actual situation, thus achieving the diagnosis of the bearing lubrication condition.

[0040] To verify the stability of this method under varying operating conditions, this embodiment applies the method to different speeds. The spindle speed is set to 400 r / min-1200 r / min, with a speed gradient of 100 r / min. Steps one through six are repeated to calculate the oil drain characteristic parameters SLF1 and SLF2 for the two test bearings at different speeds, and the SLF-speed curves are plotted, as shown below. Figure 8 As shown, the diagnostic results are consistent with the actual situation at different speeds, indicating that the method is robust to changing working conditions.

[0041] This invention proposes a signal analysis method for oil-deficient bearings based on frequency centroid index and minimum entropy deconvolution. This method can adaptively distinguish between bearing spalling / crack faults and bearing lubrication faults, effectively eliminating the influence of external large impacts and noise on the diagnostic results. The dimensionless oil-deficient characteristic parameter (SLF) proposed in this invention can effectively identify oil-deficient faults in rolling bearings and is stable under changing operating conditions. It is a reliable indicator for oil-deficient diagnosis and has important significance for monitoring bearing lubrication status in engineering applications.

Claims

1. A method for oil-starved bearing signal analysis based on frequency center of gravity index and minimum entropy deconvolution, comprising the following steps: Step one: Place the sensor on the bearing mounting seat to be tested to obtain the vibration / acoustic signal measurement value x(n) of the bearing under working condition, n=0, 1, …, N-1, N is the data length, and the sampling frequency is denoted as Fs; Step two: Fast Fourier transform is performed on the original signal x(n) to obtain the signal spectrum, and the frequency center index SCI of x(n) is calculated according to formula (1) (2) Raw ; where the sampling time interval Δt = 1 / Fs, is the one-sided power spectrum of the original signal x(n); Step 3: The original signal x(n) is processed by minimum entropy deconvolution (MED) to obtain the impulse component y(n) in the signal, and the kurtosis value of y(n) is calculated to determine whether there is a large interference pulse from an external source, and a threshold value T is set Kurt If the kurtosis value is greater than T Kurt , it is considered that the signal is contaminated, and the measurement should be re-performed, otherwise the next step of analysis is performed; Step four: Calculate the square envelope and the discrete Fourier transform of the pulse component y(n) to obtain the envelope spectrum, check whether there is any bearing fault characteristic frequency BCF in the envelope spectrum, if there is, it is determined that the bearing has spalling or crack failure; if not, it is determined that the component is aperiodic impact, i.e. random impact, and the next step is performed; Step five: the fast Fourier transform of the pulse component y(n) is carried out to obtain the signal spectrum, and the frequency center index SCI of y(n) is calculated MED , the frequency center index SCI of the original signal is known in step two Raw , the oil starvation characteristic parameter SLF of the working bearing is calculated according to formula (3); SLF = SCI MED / SCI Raw (3) Step six: set the threshold T SLF And make a judgment, if the oil characteristic parameter SLF is greater than T SLF Then it is determined that the working bearing is in the oil starvation state, otherwise the working bearing is in the appropriate lubrication state.

Images