CN109029999B - Rolling bearing fault diagnosis method based on enhanced modulation bispectrum analysis - Google Patents

Abstract

The invention discloses a rolling bearing fault diagnosis method based on enhanced modulation bispectrum analysis, which is provided aiming at the defect that the modulation bispectrum analysis can only inhibit Gaussian noise theoretically but can not take the role of non-Gaussian noise. The method specifically comprises the following steps: firstly, measuring a vibration signal of a detected rolling bearing through a vibration sensor; secondly, performing noise reduction processing on the obtained vibration signal by using an AR model to obtain a noise reduction vibration signal; and finally, carrying out MSB separation modulation on the noise reduction vibration signal, and extracting fault characteristic frequency. The rolling bearing fault diagnosis method for enhancing modulation bispectrum analysis can effectively extract weak characteristic information of a faulty bearing in strong background noise, and is favorable for finding early faults of the bearing.

Description

Rolling bearing fault diagnosis method based on enhanced modulation bispectrum analysis

Technical Field

The invention relates to the technical field of state monitoring and fault diagnosis of mechanical equipment, in particular to a rolling bearing fault diagnosis method based on enhanced modulation bispectrum analysis.

Background

Rolling bearings are the most widely used mechanical parts in rotating machinery, and are among the most vulnerable elements. Among the rotary mechanical vibration signals, a large number of signals are non-stationary and non-gaussian distributed signals, especially in the event of a fault. However, conventional power spectral analysis and time-frequency analysis cannot reflect phase information between frequency components, and usually cannot process non-minimum phase systems and non-gaussian signals, while modulation bispectrum analysis (MSB) is a powerful tool for analyzing non-stationary and non-gaussian signals. The MSB compensates for the disadvantage that the second order statistic contains no phase information and has a modulating property, so that useful fault signature information is more easily obtained with a modulated dual spectrum vibration signal. However, the MSB theoretically can completely suppress gaussian noise, and has no effect on the absence of non-gaussian noise, and the presence of these non-gaussian noise interferes with the higher-order spectrum of the signal, thereby adversely affecting the extraction and analysis of fault features. The mechanical fault signal often contains various noises, the signal-to-noise ratio of the signal is generally low, and particularly when a machine has an early fault, the fault signal is very weak, how to effectively extract fault characteristic information from a strong noise background directly influences the accuracy of fault diagnosis and the reliability of early fault prediction.

Disclosure of Invention

In order to solve the problems, an Autoregressive (AR) model and an MSB are combined, and a rolling bearing fault diagnosis method based on enhanced modulation bispectrum analysis is provided. The research idea is derived from respective characteristics of two signal analysis methods, the AR model can effectively process non-Gaussian noise existing in the signals, and MSB analysis inhibits the Gaussian noise.

The technical scheme for solving the technical problems is that the rolling bearing fault diagnosis method based on the enhanced modulation bispectrum analysis is designed, and the method specifically comprises the following steps:

the method comprises the following steps: measuring a vibration signal of the detected rolling bearing through a vibration sensor;

step two: carrying out noise reduction treatment on the obtained vibration signal by using an AR model to obtain a noise reduction vibration signal x (t);

step three: MSB separation modulation components are carried out on the noise reduction vibration signal x (t), and fault characteristic frequency is extracted;

compared with the prior art, the invention has the beneficial effects that:

(1) in view of the disadvantage of the MSB, which has good insensitivity to gaussian noise but cannot avoid interference of non-gaussian noise, AR models are used to improve the performance of the MSB.

(2) Considering that the order of the AR model directly influences the noise reduction performance of the AR, the optimal order of the AR model is determined in a self-adaptive mode by utilizing the maximum kurtosis principle.

(3) Experimental analysis results show that the enhanced modulation bispectrum analysis method has superior performance superior to MSB in the aspect of extracting fault characteristics, and has feasibility and effectiveness for fault diagnosis of the rolling bearing.

Drawings

FIG. 1 is a time domain waveform diagram of a vibration signal of an inner ring of a rolling bearing of embodiment 1;

FIG. 2 is a frequency domain graph of a vibration signal of an inner ring of the rolling bearing of embodiment 1;

fig. 3 is an AR model of the vibration signal of the inner ring of the rolling bearing of embodiment 1 in the optimal order and the corresponding maximum kurtosis value;

fig. 4 is a frequency domain diagram of a vibration signal of an inner ring of a rolling bearing of embodiment 1 obtained by the rolling bearing fault diagnosis method of the present invention using the enhanced modulation bispectrum analysis;

fig. 5 is a frequency domain graph in which the MSB is used as the vibration signal of the inner ring of the rolling bearing of example 1.

Detailed Description

Specific examples of the present invention are given below. The specific examples are only intended to illustrate the invention in further detail and do not limit the scope of protection of the claims of the present application.

The invention provides a rolling bearing fault diagnosis method based on enhanced modulation bispectrum analysis, which comprises the following specific steps:

the method comprises the following steps: measuring a vibration signal of the detected rolling bearing through a vibration sensor;

step two: carrying out noise reduction treatment on the obtained vibration signal by using an AR model to obtain a noise reduction vibration signal x (t);

the second step specifically comprises the following steps:

step 101: determining a suitable order range for the AR model is in fact possible if the general signal is taken to within 100.

Step 102: determining the weighting a of the AR model under the corresponding order by using the least square method_i(i ═ 1, 2.. times, p), preprocessing the vibration signal by using the parameter and an AR model corresponding to the order, and then calculating to obtain a corresponding kurtosis value;

step 103: comparing kurtosis values of vibration signals obtained by calculation under different orders to find out the maximum kurtosis value, wherein the corresponding order is the optimal order to be determined, and further obtaining a noise reduction vibration signal x (t);

step three: MSB separation modulation components are carried out on the noise reduction vibration signal x (t), and fault characteristic frequency is extracted;

the third step specifically comprises the following steps:

step 104: in the frequency domain, the MSB of the noise reduction vibration signal x (t) represented in the form of a discrete fourier transform x (f) may be defined as:

B_MS(f_c，f_x)＝E＜X(f_c+f_x)X(f_c-f_x)X^*(f_c)X^*(f_c)＞

wherein B is_MS(f_c，f_x) Representing the bispectrum of the signal x (t), E<>Indicates expectation, f_cTo modulate frequency, f_xIs the carrier frequency, (f)_c+f_x) And (f)_c-f_x) Upper and lower sideband frequencies, respectively.

Step 105: the MSB obtained in step 104 is improved by modifying the carrier frequency f by removing substantial influence_cComponents to more accurately quantify sideband amplitude. The improved MSB is MSB-SE, and is defined as follows:

wherein B is_MS(f_c0) represents f_xSquare power spectrum at 0.

Step 106: is calculated at f_xAverage value of MSB in increment direction to obtain f_cSlicing:

wherein Δ f represents f_xThe resolution of (2).

Step 107: calculating the average value of a plurality of optimal MSB slices to obtain the fault characteristic frequency of the rolling bearing, wherein the fault characteristic frequency is expressed as:

wherein N is selected f_cTotal number of slices.

Example 1

The embodiment provides a rolling bearing fault diagnosis method based on enhanced modulation bispectrum analysis, which comprises the following steps:

the first step is as follows: the vibration sensor is used for measuring vibration signals of the inner ring of the rolling bearing, the sampling frequency of the vibration signals is 96kHz, the sampling length is 2880000, and the fault frequency of the outer ring of the bearing is 65.17 Hz. As shown in fig. 1 and 2, the waveform diagram and the frequency domain diagram of the vibration signal show that a large amount of noise exists and the component of the failure characteristic frequency cannot be extracted.

The second step is that: adaptively determining the optimal order of the AR model by using the kurtosis maximization principle, as shown in FIG. 3; performing AR model noise reduction on the vibration signal by selecting the order of the optimal AR model to obtain a noise reduction vibration signal;

the third step: and carrying out MSB separation modulation on the noise-reduction vibration signal, extracting fault characteristic frequency to obtain a frequency domain diagram as shown in figure 4, wherein the main frequency is multiple frequency of 65.17Hz and the like, and the frequency is matched with the calculated outer ring fault characteristic frequency, so that the fault characteristic information of the outer ring of the rolling bearing is accurately extracted.

In order to fully prove the superiority of the rolling bearing fault diagnosis based on the enhanced modulation bispectrum analysis of the present invention, the vibration signals of the inner ring of the rolling bearing in example 1 were compared using MSB. The resulting structure using the MSB is shown in fig. 5. As can be seen from fig. 5, the spectrum is mixed with noise and the effects of harmonics are still present. The method designed by the invention can obtain more accurate results in the diagnosis of the rolling bearing fault, and is suitable for popularization and application.

Nothing in this specification is said to apply to the prior art.

Claims (1)

1. The rolling bearing fault diagnosis method based on the enhanced modulation bispectrum analysis is characterized by comprising the following specific steps of:

the method comprises the following steps: measuring a vibration signal of the detected rolling bearing through a vibration sensor;

step two: carrying out noise reduction treatment on the obtained vibration signal by using an AR model to obtain a noise reduction vibration signal x (t);

step three: MSB separation modulation components are carried out on the noise reduction vibration signal x (t), and fault characteristic frequency is extracted;

the second step specifically comprises the following steps:

step 101: determining a suitable order range of the AR model;

step 102: determining the weighting a of the AR model under the corresponding order by using the least square method_iThe vibration signal is preprocessed by utilizing the parameter and an AR model corresponding to the order, and then a corresponding kurtosis value is obtained through calculation;

step 103: comparing kurtosis values of vibration signals obtained by calculation under different orders to find out the maximum kurtosis value, wherein the corresponding order is the optimal order to be determined, and further obtaining a noise reduction vibration signal x (t);

the third step specifically comprises the following steps:

step 104: in the frequency domain, the MSB of the noise reduction vibration signal x (t) represented in the form of a discrete fourier transform x (f) may be defined as:

B_MS(f_c,f_x)＝E＜X(f_c+f_x)X(f_c-f_x)X*(f_c)X*(f_c)＞

wherein B is_MS(f_c,f_x) Representing the bispectrum of the signal x (t), E<>Indicates expectation, f_cTo modulate frequency, f_xIs the carrier frequency, (f)_c+f_x) And (f)_c-f_x) Upper and lower sideband frequencies, respectively;

step 105: the MSB obtained in step 104 is improved by modifying the carrier frequency f by removing substantial influence_cComponents to more accurately quantize sideband amplitude; the improved MSB is MSB-SE, and is defined as follows:

wherein B is_MS(f_c0) represents f_xSquared power spectrum when 0;

step 106: is calculated at f_xAverage value of MSB in increment direction to obtain f_cSlicing:

wherein Δ f represents f_xThe resolution of (a);

step 107: calculating the average value of a plurality of optimal MSB slices to obtain the fault characteristic frequency of the rolling bearing, wherein the fault characteristic frequency is expressed as:

wherein f is_x> 0, N is selected f_cTotal number of slices.

Images