CN104330258A - Method for identifying grey relational degree of rolling bearing fault based on characteristic parameters - Google Patents

Abstract

The invention discloses a method for identifying a grey relational degree of a rolling bearing fault based on characteristic parameters. The method comprises the step of acquiring a rolling bearing vibration signal and forming a reconstructed signal, the step of extracting the characteristic parameters from the reconstructed signal, and the step of analyzing the characteristic parameters by use of the theoretical grey relational degree and outputting an analysis result. The method for identifying the grey relational degree of the rolling bearing fault based on the characteristic parameters is capable of monitoring the operative conditions of a rolling bearing under relatively strong background noise, thereby achieving the purpose of discovering and determining the rolling bearing fault and avoiding serious faults of the mechanical equipment.

Description

A kind of rolling bearing fault grey relational grade discrimination method of feature based parameter

Technical field

The invention belongs to rolling bearing fault diagnosis technical field, particularly a kind of rolling bearing fault grey relational grade discrimination method based on LMD and SVD characteristic parameter.

Background technology

Modernization commercial production more and more maximizes, robotization, complicated; the particularly industrial sector such as petrochemical industry, metallurgy, mining; equipment investment is large and need continuous seepage operation, and mechanical equipment fault is shut down may cause heavy economic losses, even jeopardizes the personal safety of equipment operator.Rolling bearing, as parts important in rotating machinery, is also one of important source of trouble of plant equipment.Therefore, the condition monitoring and fault diagnosis technical method of research rolling bearing is for the operational efficiency and the maintenance usefulness that improve plant equipment, and the personnel property loss of avoiding has important practical significance.

At present, the application of various vibration signal processing methods in mechanical fault diagnosis, facilitates the development of fault diagnosis technology to a great extent.Fault Diagnosis of Roller Bearings is varied as methods such as time domain charactreristic parameter method, scramble spectrometry, envelope spectrometry, wavelet transformation, Wigner-Ville distribution, EMD decomposition.But these methods have respective limitation, indexs different in time domain charactreristic parameter method only differentiates comparatively effective to specific bearing defect, the scramble spectrometry in frequency domain and envelope spectrometry are difficult to the bearing defect frequency under the complicated states such as discovery identification strong noise background.Wavelet transformation needs to wait method choice wavelet basis function by rule of thumb, and adaptivity is poor.Empirical mode decomposition (EMD) method proposed in recent years and a kind of signal Time-Frequency Analysis Method developed rapidly, be suitable for processing non-stationary, nonlinear properties, there is very strong adaptivity, but EMD decomposition method also also exists end effect, modal overlap, the iterative loop shortcoming such as often.

Summary of the invention

Fundamental purpose of the present invention is the rolling bearing fault grey relational grade discrimination method providing a kind of feature based parameter, utilize the method to reach to monitor rolling bearing operating condition under comparatively strong background noise, to find and to judge rolling bearing fault, plant equipment is avoided to occur the object of comparatively catastrophic failure.

The invention provides a kind of rolling bearing fault grey relational grade discrimination method of feature based parameter, comprising:

The step of bearing vibration signal acquisition and formation reconstruction signal;

The step of characteristic parameter is extracted from described reconstruction signal;

Utilize characteristic parameter described in gray theory correlation analysis and export the step of analysis result.

Further, described vibration signal includes rolling bearing normal operation signal, inner ring defect fault-signal, outer ring defect fault-signal and rolling body defect fault-signal.

Further, the step of described bearing vibration signal acquisition and formation reconstruction signal comprises:

Gather the step of vibration signal;

Hilbert conversion is carried out to described vibration signal and forms the step of described amplitude signal and phase signal;

SVD singular value decomposition method is utilized to carry out noise reduction process to described amplitude signal and described phase signal respectively and the step of phase signal after amplitude signal and noise reduction after forming noise reduction;

The step of reconstruction signal is formed in conjunction with phase signal after amplitude signal after described noise reduction and described noise reduction.

Further, to described amplitude signal and described phase signal, noise reduction process is carried out respectively to the described SVD of utilization singular value decomposition method and after forming noise reduction after amplitude signal and noise reduction the step of phase signal determine the additional step of noise reduction order.

Further, the reconstruction signal formula that after noise reduction described in described combination, after amplitude signal and described noise reduction, phase signal forms the step of reconstruction signal is:

Wherein A _c(t) and be respectively the amplitude after noise reduction and phase signal, x _ct () is reconstruction signal.

Further, the described step extracting characteristic parameter from described reconstruction signal comprises:

Part mean decomposition method is utilized to carry out described reconstruction signal decomposing the step obtaining envelope pure frequency modulation PF component;

By the step of pure for described envelope frequency modulation PF component composition PF matrix;

PF matrix described in SVD svd is utilized to extract the step of feature singular value;

Statistical theory processing feature singular value is utilized to obtain the step of characteristic parameter.

Further, described envelope pure frequency modulation PF component is the pure frequency modulation components of front 5 intrinsic envelopes.

Further, described characteristic parameter includes the maximal value of singular value, singular value average, the singular value pulse factor, singular entropy and singular value standard deviation.

Further, described utilize characteristic parameter described in gray theory correlation analysis and export fault in the step of analysis result distinguish that method for distinguishing is for grey Relational Analysis Method.

Further, the formula of described grey Relational Analysis Method is as follows:

${\mathrm{\ζ}}_i\left(k\right)=\frac{\underset{s}{\min }\underset{t}{\min }|x_0\left(t\right)-x_s\left(t\right)|+\mathrm{\ρ}\underset{s}{\max }\underset{t}{\max }|x_0\left(t\right)-x_s\left(t\right)|}{|x_0\left(k\right)-x_i\left(k\right)|+\mathrm{\ρ}\underset{s}{\max }\underset{t}{\max }|x_0\left(t\right)-x_s\left(t\right)|}$

Wherein x ₀t () is reference time array, x _i(t) for comparing time series, ζ _ik () is for comparing time series x _it () is to reference time array x _it (), at the correlation coefficient in k moment, ρ ∈ [0,1] gets 0.5 usually for resolution ratio.Be calculated as follows and compare time series x _it () is to reference time array x ₀the degree of association of (t):

$r_i=\frac{1}{n}\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{\ζ}}_i\left(k\right)$

Beneficial effect of the present invention is, a kind of rolling bearing fault grey relational grade discrimination method of feature based parameter is provided, the method can not effective identification bearing defect frequency under comparatively strong background noise for traditional Fault Diagnosis of Roller Bearings, propose Hilbert conversion and SVD svd associating noise-reduction method and local mean value decomposition method and singular value decomposition method to combine the method for extraction bearing fault characteristics, and utilize gray theory correlation analysis identification bearing fault classification, abundant and perfect to a certain extent method for diagnosing faults; Singular value decomposition method after the selection of noise reduction order, to signals and associated noises noise reduction, effectively can improve the signal to noise ratio (S/N ratio) of signal; Can fast and effeciently extract Rolling Bearing Fault Character information by the method after combining, and complete the identification for rolling bearing rolling body defect fault, inner ring defect fault, outer ring defect fault and normal four class bearing states.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of the rolling bearing fault grey relational grade discrimination method of embodiment of the present invention feature based parameter;

Fig. 2 is the particular flow sheet of the rolling bearing fault grey relational grade discrimination method of a kind of feature based parameter of the embodiment of the present invention;

Fig. 3 a is the bearing vibration signal time-domain diagram that the embodiment of the present invention has outer ring fault;

Fig. 3 b is that the rolling bearing that the embodiment of the present invention has an outer ring fault adds the vibration signal time-domain diagram after making an uproar;

Fig. 4 a is the outer ring faulty bearing vibration signal time-domain diagram after the embodiment of the present invention adopts the method for the invention to carry out noise reduction process;

Fig. 4 b is the outer ring faulty bearing vibration signal envelope spectrogram after embodiment of the present invention noise reduction;

Fig. 5 is the partial enlarged drawing of the outer ring faulty bearing vibration signal envelope spectrogram after embodiment of the present invention noise reduction;

Fig. 6 is the outer ring faulty bearing vibration signal LMD exploded view after embodiment of the present invention noise reduction;

Fig. 7 is the grey relational grade identification result figure of embodiment of the present invention experiment bearing inner race fault, outer ring fault, rolling body fault and normal four kinds of states.

Embodiment

Hereafter will describe embodiments of the invention in detail by reference to the accompanying drawings.It should be noted that the combination of technical characteristic or the technical characteristic described in following embodiment should not be considered to isolated, they mutually can be combined and be combined with each other thus reach better technique effect.

Fig. 1 is the process flow diagram of the rolling bearing fault grey relational grade discrimination method of embodiment of the present invention feature based parameter.

The embodiment of the present invention provides a kind of rolling bearing fault grey relational grade discrimination method of feature based parameter, comprising:

The step of bearing vibration signal acquisition and formation reconstruction signal;

The step of characteristic parameter is extracted from described reconstruction signal;

Utilize characteristic parameter described in gray theory correlation analysis and export the step of analysis result.

The embodiment of the present invention, described vibration signal includes rolling bearing normal operation signal, inner ring defect fault-signal, outer ring defect fault-signal and rolling body defect fault-signal.

As shown in Figure 2, a kind of rolling bearing fault grey relational grade discrimination method based on LMD and SVD characteristic parameter described in invention comprises the following steps:

The embodiment of the present invention, the step of described bearing vibration signal acquisition and formation reconstruction signal comprises:

Gather the step of vibration signal; In addition the embodiment of the present invention is in order to simulate comparatively strong background noise, suitably adds the white Gaussian noise of some strength in the bearing vibration signal gathered.

Hilbert conversion is carried out to described vibration signal and forms the step of described amplitude signal and phase signal; Hilbert conversion is carried out to noisy vibration signal, obtains amplitude and the phase time sequence of noisy vibration signal, Hilbert transform is carried out for Vibration Signal Time Series x (t) comprising noise:

$\hat{x}\left(t\right)=H\left[x\left(t\right)\right]=\frac{1}{\mathrm{\π}}{\∫}_{-\∞}^{\∞}\frac{x\left(\mathrm{\τ}\right)}{t-\mathrm{\τ}}\mathrm{d\τ}=x\left(t\right)*\frac{1}{\mathrm{\πt}}---\left(1\right)$

Then the analytic signal of x (t) is:

Amplitude and the phase place of analytic signal are respectively

$A\left(t\right)=\sqrt{x^2\left(t\right)+{\hat{x}}^2\left(t\right)}---\left(3\right)$

SVD singular value decomposition method is utilized to carry out noise reduction process to described amplitude signal and described phase signal respectively and the step of phase signal after amplitude signal and noise reduction after forming noise reduction; Amplitude after svd SVD noise reduction and phase signal are designated as A respectively _c(t) and shape is constructed respectively as D for above-mentioned amplitude and phase signal time series _msvd matrix

$D_m={\left(\begin{array}{cccc}{x}_1& x_2& L& x_n\\ x_{1\×\mathrm{\τ}+1}& x_{1\×\mathrm{\τ}+2}& L& x_{1\×\mathrm{\τ}+n}\\ M& M& & M\\ x_{(m-1)\×\mathrm{\τ}+1}& x_{(m-1)\×\mathrm{\τ}+2}& L& x_{(m-1)\×\mathrm{\τ}+n}\end{array}\right)}_{m\×n}---\left(5\right)$

By signal time sequence matrix D _mcarry out svd

D _m＝USV ^T(6)

U is m × m matrix, and V is n × n matrix, and UU ^t=I, VV ^t=I, S are that m × n ties up diagonal matrix, and diagonal entry is λ ₁, λ ₂, K, λ _k, the order of A is k, and k≤min (m, n), λ ₁>=λ ₂>=K>=λ _kbe called the singular value of matrix A.

To described amplitude signal and described phase signal, noise reduction process is carried out respectively to the described SVD of utilization singular value decomposition method and after forming noise reduction after amplitude signal and noise reduction the step of phase signal determine the additional step of noise reduction order; Singular value amplitude fluctuations for noisy vibration signal then gets a front p singular value by formula (6) reconstruction signal and settling signal noise reduction relatively gently after p (k<p) is individual.Wherein p is noise reduction order, adopts equation of the ecentre commercial law to choose noise reduction order here.

$Z_i=\frac{{\mathrm{\&sigma;}}_{i-1}-{\mathrm{\&sigma;}}_{i+1}}{2},i=\mathrm{2,3},L,k-1---\left(7\right)$

I is singular value sequence number, σ _ibe i-th singular value

Converted can be obtained by analytic signal (2) formula:

The step of reconstruction signal is formed in conjunction with phase signal after amplitude signal after described noise reduction and described noise reduction; In conjunction with (2) and (8) known reconstruction signal x _ct () can be expressed as:

The embodiment of the present invention, the described step extracting characteristic parameter from described reconstruction signal comprises:

Part mean decomposition method is utilized to carry out described reconstruction signal decomposing the step obtaining envelope pure frequency modulation PF component; To reconstruction signal x _ct () carries out local mean value decomposition, first determine signal x _ct the Local modulus maxima on () and local minizing point, adopt Cubic Spline Method to obtain upper and lower envelope u ₁(t) and v ₁(t).Calculate local mean value function m respectively ₁₁(t) and envelope estimation function a ₁₁t () is as follows:

$m_{11}\left(t\right)=\frac{u_1\left(t\right)+v_1\left(t\right)}{2}---\left(10\right)$

$a_{11}\left(t\right)=\frac{|u_1\left(t\right)-v_1\left(t\right)|}{2}---\left(11\right)$

By local mean value function from reconstruction signal x _cresidual signal h is separated to obtain in (t) ₁₁t (), uses h ₁₁t () is divided by envelope estimation function a ₁₁t () is to h ₁₁t () is carried out demodulation and is obtained

h ₁₁(t)＝x _c(t)-m ₁₁(t) (12)

s ₁₁(t)＝h ₁₁(t)/a ₁₁(t) (13)

S is calculated by formula (11) ₁₁the envelope estimation function a of (t) ₁₂t (), if a ₁₂(t)=1, then s ₁₁t () is a pure FM signal, otherwise iteration formula (10) ~ (13) are until s _1nt () is a pure FM signal.Iterative process is as follows:

$\left\{\begin{array}{c}{h}_{11}\left(t\right)=x_c\left(t\right)-m_{11}\left(t\right)\\ h_{12}\left(t\right)=s_{11}\left(t\right)-m_{12}\left(t\right)\\ M\\ h_{1n}\left(t\right)=s_{1(n-1)}\left(t\right)-m_{1n}\left(t\right)\end{array}\right.---\left(14\right)$

$\left\{\begin{array}{c}{s}_{11}\left(t\right)=h_{11}\left(t\right)/a_{11}\left(t\right)\\ s_{12}\left(t\right)=h_{12}\left(t\right)/a_{12}\left(t\right)\\ M\\ s_{1n}\left(t\right)=h_{1n}\left(t\right)/a_{1n}\left(t\right)\end{array}\right.---\left(15\right)$

Usually, theoretic iteration intermediate value condition is

$\underset{n\&RightArrow;\&infin;}{\lim }{a}_{1n}\left(t\right)=1---\left(16\right)$

The envelope estimation function obtained in above-mentioned iterative process is all multiplied and can obtains envelope signal

$a_1\left(t\right)=a_{11}\left(t\right)a_{12}\left(t\right)L,a_{1n}\left(t\right)=\underset{i=1}{\overset{n}{\mathrm{\&Pi;}}}{a}_{1i}\left(t\right)---\left(17\right)$

The envelope signal a that upper step is obtained ₁(t) and pure FM signal s _1nt () is multiplied just can obtain reconstruction signal x _ct first PF component of () is

pf ₁＝a ₁(t)gs _1n(t) (18)

Reconstruction signal x _ct () deducts first PF component can obtain new signal c ₁t (), repeats above-mentioned iterative step to it until residual signal is a monotonic quantity.

$\left\{\begin{array}{c}{c}_1\left(t\right)=x_c\left(t\right)-{\mathrm{pf}}_1\left(t\right)\\ c_2\left(t\right)=c_1\left(t\right)-{\mathrm{pf}}_2\left(t\right)\\ M\\ c_n\left(t\right)=c_{n-1}\left(t\right)-{\mathrm{pf}}_n\left(t\right)\end{array}---\left(19\right)\right.$

Superpose each PF component above-mentioned and can obtain reconstruction signal x _c(t) be

$x_c\left(t\right)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{pf}}_i\left(t\right)+c_n\left(t\right)---\left(20\right)$

Adopt LMD method to after the vibration signal after SVD noise reduction decomposes, several PF components obtained.

By the step of pure for described envelope frequency modulation PF component composition PF matrix; Described some PF components contain different frequency contents respectively, choose front 5 PF components composition initial characteristics vector matrix J

J＝[pf ₁ ^T(t),pf ₂ ^T(t),L,pf ₅ ^T(t)] ^T(21)

PF matrix described in SVD svd is utilized to extract the step of feature singular value;

Statistical theory processing feature singular value is utilized to obtain the step of characteristic parameter; Use statistical theory according to the maximal value λ of corresponding formulae discovery singular value _max, singular value average λ _ave, pulse factor I, singular entropy S _sum, standard deviation S _td, with the proper vector p=[λ obtained _max, λ _ave, I, S _sum, S _td].Computing formula is as follows respectively:

A. singular value maximal value λ _max: λ _max=max (λ ₁, λ ₂, L λ _n) (22)

B. singular value average ${\mathrm{\&lambda;}}_{\mathrm{ave}}:{\mathrm{\&lambda;}}_{\mathrm{ave}}=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_i---\left(23\right)$

C. singular value pulse factor I:I=λ ₁/ λ _ave(24)

D. singular entropy $S_{\mathrm{sum}}:\left\{\begin{array}{c}{S}_{\mathrm{sum}}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(-p_i\lg p_i)\\ p_i={\mathrm{\&lambda;}}_i/\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_i\end{array}\right.---\left(25\right)$

E. singular value standard deviation $S_{\mathrm{td}}:S_{\mathrm{td}}=\sqrt{\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{({\mathrm{\&lambda;}}_i-{\mathrm{\&lambda;}}_{\mathrm{ave}})}^2}---\left(26\right)$

The embodiment of the present invention, described to utilize characteristic parameter described in gray theory correlation analysis and export fault discriminating conduct in the step of analysis result be grey Relational Analysis Method, and the formula of described grey Relational Analysis Method is as follows

${\mathrm{\&zeta;}}_i\left(k\right)=\frac{\underset{s}{\min }\underset{t}{\min }|x_0\left(t\right)-x_s\left(t\right)|+\mathrm{\&rho;}\underset{s}{\max }\underset{t}{\max }|x_0\left(t\right)-x_s\left(t\right)|}{|x_0\left(k\right)-x_i\left(k\right)|+\mathrm{\&rho;}\underset{s}{\max }\underset{t}{\max }|x_0\left(t\right)-x_s\left(t\right)|}---\left(27\right)$

Wherein x ₀t () is reference time array, x _i(t) for comparing time series, ζ _ik () is for comparing time series x _it () is to reference time array x ₀t (), at the correlation coefficient in k moment, ρ ∈ [0,1] gets 0.5 usually for resolution ratio.Be calculated as follows and compare time series x _it () is to reference time array x ₀the degree of association of (t)

$r_i=\frac{1}{n}\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&zeta;}}_i\left(k\right)---\left(28\right)$

In rolling bearing fault identification, eigenvectors matrix is divided into two parts by 1:1 by us, and a part asks its mean value as master sample, the x namely in above formula ₀(t), the x in the similar above formula of another part _it (), as test sample book, carries out fault identification.

The embodiment of the present invention, integrated use said method carries out a specific embodiment of rolling bearing fault diagnosis.

The rolling bearing model that this specific embodiment uses is 6205-2RS JEM SKF, spark erosion technique is used to arrange Single Point of Faliure on bearing, fault diameter is 0.007 inch of (about 0.178mm) motor speed is 1797rpm (29.95Hz), sample frequency 12k Hz, uses the bearing data near motor drive terminal.

Here for bearing outer ring fault, noisy vibration signal noise reduction preprocessing process is described.Through calculating outer ring, faulty bearing characteristic frequency is 107.4Hz, and the number of data points that outer ring faulty bearing uses is 2400.As shown in Figure 3 a, Fig. 3 b is the outer ring faulty bearing status signal time-domain diagram adding-3dB white Gaussian noise to the time domain beamformer of bearing outer ring fault-signal.Fig. 4 a is the outer ring faulty bearing vibration signal time-domain diagram after adopting Hilbert of the present invention conversion and SVD svd integrated processes to carry out noise reduction process, and comparison diagram 3 and Fig. 4 a can find out after the method process, effectively eliminate noise.Hilbert conversion is carried out to the outer ring faulty bearing vibration signal after noise reduction and asks envelope, and Fast Fourier Transform (FFT) is carried out to envelope signal obtain envelope spectrum signal as shown in Figure 4 b.Due to experiment sample frequency 12kHz, the number of data points of use is 2400 points, therefore frequency resolution is 5Hz.Can find out from Fig. 5 envelope frequency spectrum figure and occur obvious peak value at 30Hz, 110Hz, 215Hz, 430Hz, 540Hz and 645Hz.This and experiment shaft frequency 29.95Hz, outer ring fault characteristic frequency 107.4Hz and 2,4,5, the value of 6 frequencys multiplication is close, frequency departure part may be relevant with frequency resolution and other influence factor.The frequecy characteristic value that housing washer fault envelope spectrogram occurs also shows to adopt noise-reduction method of the present invention can effectively retain Rolling Bearing Fault Character information, for process is prepared further.Fig. 6 be noise reduction after outer ring faulty bearing vibration signal LMD exploded view; Decomposed the front 5 layers of PF component composition initial vector matrix extracting outer ring faulty bearing vibration signal by LMD, utilize singular value decomposition method obtain characterizing the feature singular value of failure message and utilize statistical theory process singular value features amount.Singular value features amount obtained above is input to grey relation model and carries out rolling bearing fault identification.From rolling bearing experiment obtain bearing outer ring fault, inner ring fault, roller fault and normal four kinds of status signals, each 20 groups of every type signal, get wherein 10 groups as master sample, all the other 10 groups is test sample book, often organizes 2400 data points.Use rolling bearing fault discrimination method of the present invention to carry out fault diagnosis, result as shown in Figure 7.Wherein 1-10 group is bearing inner race fault test data, and 11-20 group is bearing outer ring fault test data, and 21-30 group is rolling body fault test data, and 31-40 group is bearing normal condition test data.Inner ring fault only has the 5th group to be identified as outer ring fault, group is only had to be that 19,20 liang of groups are identified as inner ring fault in the fault of outer ring, all the other Rolling Bearing Status identifications are good, and Classification and Identification accuracy reaches 92.5%, show that the method for the invention can effective identification rolling bearing fault type.

The invention provides a kind of rolling bearing fault grey relational grade discrimination method of feature based parameter, the method can not effective identification bearing defect frequency under comparatively strong background noise for traditional Fault Diagnosis of Roller Bearings, propose Hilbert conversion and SVD svd associating noise-reduction method and local mean value decomposition method and singular value decomposition method to combine the method for extraction bearing fault characteristics, and utilize gray theory correlation analysis identification bearing fault classification, abundant and perfect to a certain extent method for diagnosing faults; Singular value decomposition method after the selection of noise reduction order, to signals and associated noises noise reduction, effectively can improve the signal to noise ratio (S/N ratio) of signal; Can fast and effeciently extract Rolling Bearing Fault Character information by the method after combining, and complete the identification for rolling bearing rolling body defect fault, inner ring defect fault, outer ring defect fault and normal four class bearing states.

Although give some embodiments of the present invention, it will be understood by those of skill in the art that without departing from the spirit of the invention herein, can change embodiment herein.Above-described embodiment is exemplary, should using embodiment herein as the restriction of interest field of the present invention.

Claims (10)

1. a rolling bearing fault grey relational grade discrimination method for feature based parameter, is characterized in that, comprising:

The step of bearing vibration signal acquisition and formation reconstruction signal;

The step of characteristic parameter is extracted from described reconstruction signal;

Utilize characteristic parameter described in gray theory correlation analysis and export the step of analysis result.

2. the rolling bearing fault grey relational grade discrimination method of feature based parameter as claimed in claim 1, it is characterized in that, described vibration signal includes rolling bearing normal operation signal, inner ring defect fault-signal, outer ring defect fault-signal and rolling body defect fault-signal.

3. the rolling bearing fault grey relational grade discrimination method of feature based parameter as claimed in claim 1, is characterized in that, the step of described bearing vibration signal acquisition and formation reconstruction signal comprises:

Gather the step of vibration signal;

Hilbert conversion is carried out to described vibration signal and forms the step of described amplitude signal and phase signal;

SVD singular value decomposition method is utilized to carry out noise reduction process to described amplitude signal and described phase signal respectively and the step of phase signal after amplitude signal and noise reduction after forming noise reduction;

The step of reconstruction signal is formed in conjunction with phase signal after amplitude signal after described noise reduction and described noise reduction.

4. the rolling bearing fault grey relational grade discrimination method of feature based parameter as claimed in claim 3, it is characterized in that, to described amplitude signal and described phase signal, noise reduction process is carried out respectively to the described SVD of utilization singular value decomposition method and after forming noise reduction after amplitude signal and noise reduction the step of phase signal determine the additional step of noise reduction order.

5. the rolling bearing fault grey relational grade discrimination method of feature based parameter as claimed in claim 3, is characterized in that, the reconstruction signal formula that after noise reduction described in described combination, after amplitude signal and described noise reduction, phase signal forms the step of reconstruction signal is:

Wherein A _c(t) and be respectively the amplitude after noise reduction and phase signal, x _ct () is reconstruction signal.

6. the rolling bearing fault grey relational grade discrimination method of feature based parameter as claimed in claim 1, it is characterized in that, the described step extracting characteristic parameter from described reconstruction signal comprises:

Part mean decomposition method is utilized to carry out described reconstruction signal decomposing the step obtaining envelope pure frequency modulation PF component;

By the step of pure for described envelope frequency modulation PF component composition PF matrix;

PF matrix described in SVD svd is utilized to extract the step of feature singular value;

Statistical theory processing feature singular value is utilized to obtain the step of characteristic parameter.

7. the rolling bearing fault grey relational grade discrimination method of feature based parameter as claimed in claim 6, it is characterized in that, described envelope pure frequency modulation PF component is the pure frequency modulation components of front 5 intrinsic envelopes.

8. the rolling bearing fault grey relational grade discrimination method of feature based parameter as claimed in claim 6, is characterized in that, described characteristic parameter includes the maximal value of singular value, singular value average, the singular value pulse factor, singular entropy and singular value standard deviation.

9. the rolling bearing fault grey relational grade discrimination method of feature based parameter as claimed in claim 1, it is characterized in that, described utilize characteristic parameter described in gray theory correlation analysis and export fault in the step of analysis result distinguish that method for distinguishing is for grey Relational Analysis Method.

10. the rolling bearing fault grey relational grade discrimination method of feature based parameter as claimed in claim 9, it is characterized in that, the formula of described grey Relational Analysis Method is as follows:

Wherein x ₀t () is reference time array, x _i(t) for comparing time series, ζ _ik () is for comparing time series x _it () is to reference time array x _it (), at the correlation coefficient in k moment, ρ ∈ [0,1] gets 0.5 usually for resolution ratio.Be calculated as follows and compare time series x _it () is to reference time array x ₀the degree of association of (t):

。