CN103900815A - Rolling bearing fault diagnosis method based on EEMD and distribution fitting testing - Google Patents

Abstract

The invention discloses a rolling bearing fault diagnosis method based on EEMD and distribution fitting testing, and relates to the field of rotary machine fault diagnosis. The method includes the steps that original signals are decomposed into a series of IMF components by means of EEMD; then each IMF component is sampled to obtain a sampling point; testing is carried out through a normal probability plot and a Jarque-Bera test, and whether data accord with normal distribution or not is judged; the IMF components according with white noise characteristics are removed, and signals after other components are added and noise reduction is carried out are reserved; finally, fault diagnosis is carried out on rolling bearings by means of an envelope spectrum. According to the method, the aim of signal noise reduction can be well achieved, signal fault features can be more obvious, ultimately fault frequency is identified by means of the envelope spectrum, and fault types of the rolling bearings can be well diagnosed.

Description

A kind of Fault Diagnosis of Roller Bearings based on EEMD and fitting of distribution check

Technical field

The present invention relates to rotary machinery fault diagnosis field, in particular a kind of Fault Diagnosis of Roller Bearings based on EEMD and fitting of distribution check.

Background technology

Rolling bearing is widely used in rotating machinery, and it is also the critical component of large rotating machinery.Due to its architectural characteristic, rolling bearing is to hold flimsy part, and in the polytype fault occurring at rotating machinery, many all have relation with the damage of rolling bearing.The fault of plant equipment 70% is vibration fault, in these vibration faults, approximately has 30% to be caused by bearing.This is because the working environment very severe of rolling bearing, once rolling bearing breaks down, will cause a series of chain reactions of whole plant equipment, brings huge potential safety hazard to plant equipment.Therefore,, in mechanical fault diagnosis technology, the fault diagnosis technology of rolling bearing is important ingredient.

The method for diagnosing faults of rolling bearing has a lot, and rolling bearing fault diagnosis technology based on vibration signal is the most frequently used, is also effective method.Vibration signal has comprised abundant plant equipment extremely or failure message, and it can reflect the status flag that plant equipment is moved.Traditional most widely used Digital Signal Processing is exactly Fourier transform.Invent FFT(Fast Fourier Transform (FFT) from nineteen sixty-five Cooley-Tu Ji, FastFourier Transformation) afterwards, Spectral Analysis Method is all widely used in every field.Wavelet transformation is the novel Digital Signal Processing growing up in the later stage eighties, and its essence is a Fourier transform that window is adjustable, and its time window and frequency window can change, and is a kind of analytical approach of time frequency localization.FFT can not provide frequency situation over time, can only provide separately the frequency distribution situation of signal, lacks the positioning function with the time.Wavelet transformation, due to wavelet basis function limited length, therefore, concerning the distribution of signal energy relative time and frequency, carries out accurate analysis to it and will produce larger difficulty.Once wavelet base is chosen, just must analyze all data to be analyzed with this wavelet basis function, therefore wavelet transformation does not possess adaptive feature.

EMD(empirical mode decomposition, empirical mode decomposition) can be limited intrinsic mode function component a complicated signal decomposition according to local time's characteristic of signal.But EMD also has shortcoming, topmost problem is exactly mode Aliasing Problem.EEMD(gathers empirical mode decomposition, ensemble empirical mode decomposition) be a kind of novel digital signal processing method, can be by a series of being decomposed into of complex fault signal adaptive intrinsic mode function components, be applicable to analysis and processing non-linear, non-stationary signal.Fitting of distribution check is whether sample value for testing one group of data is from the method for inspection of certain given distribution.

Summary of the invention

Technical matters to be solved by this invention is for the deficiencies in the prior art, and a kind of Fault Diagnosis of Roller Bearings based on EEMD and fitting of distribution check is provided.

Technical scheme of the present invention is as follows:

Based on a Fault Diagnosis of Roller Bearings for EEMD and fitting of distribution check, its step is as follows:

(1) utilize EEMD that original signal is decomposed into a series of IMF components;

(2) each IMF component is sampled, obtain sampled point; Then utilize normal probability paper figure and Jarque-Bera test to test, judge whether data meet normal distribution;

(3) remove the IMF component that meets white noise feature, retain after all the other components are added and obtain the signal after noise reduction;

(4) utilize envelope spectrum to carry out fault diagnosis to rolling bearing.

The present invention adopts EEMD to carry out the decomposition of sophisticated signal, can solve classic method and not possess adaptive shortcoming, also can alleviate the problem of the mode aliasing that EMD occurs, is more applicable for analysis pre-service non-linear, non-stationary signal; Adopt fitting of distribution check method of inspection, screen useful IMF component from the probability density characteristics of white noise IMF component, remove useless IMF component simultaneously, have good theoretical foundation; The present invention can reach the object of good signal de-noising, makes the fault signature of signal more obvious, finally utilizes envelope spectrum to identify failure-frequency, can well diagnose out the fault type of bearing.

Accompanying drawing explanation

Fig. 1 is that s (t) utilizes EMD decomposition result figure.

Fig. 2 is that s (t) utilizes EEMD decomposition result figure.

Fig. 3 is the decomposition result figure of inner ring fault-signal through EEMD.

Fig. 4 is the normal probability paper figure of the IMF component of inner ring fault-signal.

Fig. 5 is the result after inner ring fault-signal noise reduction.

Fig. 6 is inner ring fault-signal envelope spectrum.

Fig. 7 is the decomposition result figure of outer ring fault-signal through EEMD.

Fig. 8 is the normal probability paper figure of the IMF component of outer ring fault-signal.

Fig. 9 is the result after the fault-signal noise reduction of outer ring.

Figure 10 is outer ring fault-signal envelope spectrum.

Embodiment

Below in conjunction with specific embodiment, the present invention is described in detail.

Embodiment 1

1, suppose a simulate signal s (t), its formation is the stack of 7Hz sinusoidal component and Gauspuls pulse component, and Fig. 1 is the result after s (t) decomposes through EMD.

Obviously can find out from Fig. 1, clearly there is mode aliasing in IMF1 and IMF2, has been difficult to find out IMF1 and the represented physical meaning of IMF2 from figure.

2, the simulate signal s (t) in 1 is carried out to EEMD decomposition, result as shown in Figure 2.As can be seen from Figure 2, s (t) has obtained 4 IMF components and 1 surplus after decomposing through EEMD, wherein c1 and c2 represent to decompose Gauspuls pulse component out, c4 expresses 7Hz sinusoidal component well, therefore, after decomposing through EEMD, can well decomposite the composition that original signal comprises, each IMF component has the physical significance of oneself.Therefore, EEMD can solve the problem of mode aliasing well, can well reflect the essential characteristic of s (t).

Embodiment 2

Rolling bearing inner ring fault-signal comes from bearing data center of CWRU of the U.S., and wherein: sample frequency is 12000Hz, the rotating speed of motor is 1750r/min.First rolling bearing inner ring fault-signal is carried out to EEMD decomposition, obtain some intrinsic mode function components.Then according to the probability density characteristics of IMF component, utilize normal probability paper figure and Jarque-Bera method of testing to carry out the removal of choosing of useful IMF component and useless IMF component.Useful IMF component is reconstructed to the inner ring fault-signal obtaining after noise reduction, finally utilizes envelope spectrum to identify Rolling Bearing Fault Character frequency.

Fig. 3 is the result of rolling bearing inner ring fault-signal after EEMD decomposes.Fig. 4 is the normal probability paper figure of the inner ring fault IMF component that decomposes out.

For normal probability paper figure, if sampled point ("+" point in figure) approaches the dotted line in figure, can think that this group sampled point is similar to Normal Distribution, therefore, for c4, c5, c6, its characteristic distributions meets the distribution character of the IMF of white noise, therefore can set it as useless IMF component and remove.In order to improve the accuracy of check, also adopt Jarque-Bera test herein, test result is as shown in table 1.

The Jarque-Bera test result of the IMF component of table 1 inner ring fault-signal

For Jarque-Bera test, there is a H value, in the time of H=0, show the hypothesis that this group data receiver sample is normal distribution, otherwise, in the time of H=1, the hypothesis that this group data rejection sample is normal distribution.Therefore, can remove equally c4, c5, c6 component, retains all the other components, consistent with the result that normal probability paper figure obtains.Therefore, reconstruct c1, c2, c3 and c7, as shown in Figure 5, the signal after noise reduction has not only well retained the primitive character of signal, and makes fault signature more obvious.Fig. 6 is the envelope spectrum of inner ring fault-signal, and this has many failure-frequencies, the computing formula according to Bearing Analysis Theory failure-frequency: inner ring failure-frequency f _i=0.5Zf (1+dcos α D), wherein, gyro frequency is f, and roller diameter is d, and the pitch diameter of bearing is D, and roller number is Z, bearing pressure angle is α.Rolling bearing turn frequently f=29.17Hz, inner ring failure-frequency is f _i=157.9Hz, as can be seen from Figure 6, f ₁=29.3Hz, f ₂=58.6Hz, f ₃=158.2Hz, f ₄=316.5Hz.Therefore, f ₁and f ₂be approximately equal to respectively rolling bearing turn frequently and 2 times turn frequency, f ₃and f ₄be approximately equal to respectively inner ring failure-frequency and 2 times of inner ring failure-frequencies of rolling bearing.There are some errors in the calculated results and results of calculation, is because the workmanship precision of rolling bearing is not accurate enough affected.

Embodiment 3

Housing washer fault-signal still comes from bearing data center of CWRU of the U.S., and wherein: sample frequency is 12000Hz, the rotating speed of motor is 1750r/min.First housing washer fault-signal is carried out to EEMD decomposition, obtain some intrinsic mode function components.Then according to the probability density characteristics of IMF component, utilize normal probability paper figure and Jarque-Bera method of testing to carry out the removal of choosing of useful IMF component and useless IMF component.Useful IMF component is reconstructed to the outer ring fault-signal obtaining after noise reduction, finally utilizes envelope spectrum to identify Rolling Bearing Fault Character frequency.

Fig. 7 is the result of housing washer fault-signal after EEMD decomposes.Fig. 8 is the normal probability paper figure of the outer ring fault IMF component that decomposes out.Table 2 is the Jarque-Bera test result of outer ring fault IMF component.

The Jarque-Bera test result of the IMF component of table 2 outer ring fault-signal

From Fig. 8 and table 2, can find out, c1, c4, c6 and c7 do not meet the probability density characteristics of white noise IMF, therefore, retain these components, remove remaining component simultaneously.Fig. 9 is the result after the failure reconfiguration of outer ring, and Figure 10 is outer ring fault-signal envelope spectrum.Computing formula according to Bearing Analysis Theory failure-frequency: outer ring failure-frequency f _o=0.5Zf (1-dcos α D), the represented implication of correlation parameter as shown in Example 2.Rolling bearing turn frequently f=29.17Hz, outer ring failure-frequency is f _o=104.57Hz, as can be seen from Figure 10, f ₁=11.7Hz, f ₂=29.3Hz, f ₃=105.5Hz, f ₄=211Hz.Therefore, f ₁substantially equal the failure-frequency (11.6Hz) of retainer, f ₂for turning frequently of rolling bearing, f ₃and f ₄substantially equal outer ring failure-frequency and 2 times of outer ring failure-frequencies.

Should be understood that, for those of ordinary skills, can be improved according to the above description or convert, and all these improvement and conversion all should belong to the protection domain of claims of the present invention.

Claims (1)

1. the Fault Diagnosis of Roller Bearings based on EEMD and fitting of distribution check, is characterized in that, its step is as follows:

(1) utilize EEMD that original signal is decomposed into a series of IMF components;

(2) each IMF component is sampled, obtain sampled point; Then utilize normal probability paper figure and Jarque-Bera test to test, judge whether data meet normal distribution;

(3) remove the IMF component that meets white noise feature, retain after all the other components are added and obtain the signal after noise reduction;

(4) utilize envelope spectrum to carry out fault diagnosis to rolling bearing.

Images