CN104318043B - Rolling bearing vibration performance reliability variation process detection method and rolling bearing vibration performance reliability variation process detection device - Google Patents

Abstract

The invention relates to a rolling bearing vibration performance reliability variation process detection method and a rolling bearing vibration performance reliability variation process detection device. In the method provided by the invention, according to a time sequence counting process, extremely little variation intensity original information shown by bearing vibration is obtained at a short time interval; through the self-help re-sampling on the variation intensity original information, much variation intensity generating information is simulated; the generating information is processed by using a grey prediction model, and a variation intensity estimation value is obtained; a Poisson process is used for expressing a reliability function; and the variation process of the bearing vibration performance reliability is predicted in real time. The method and the device provided by the invention have the advantages that on the basis of the time sequence of the vibration information, the grey self-help principle is merged into the Poisson process, and a grey self-help Poisson method is provided for predicating the variation process of the rolling bearing performance reliability.

Description

Bearing vibration performance reliability mutation process detection method and device

Technical field

The present invention relates to a kind of bearing vibration performance reliability mutation process detection method and device.

Background technology

For guaranteeing safe operation, many systems, such as spacecraft, aircraft, bullet train and nuclear reactor etc., to rolling bearing Reliability during one's term of military service has very strict requirements.Bearing can appear before failure many fucoids as, for example vibration, temperature rise or The performances such as friction become abnormal, indicate damage or the abrasion degradation mode of Bearing inner part, therefore, real-time assessment and prediction axle Variation (change/degenerate) process of performance reliability is held, the hidden danger that fails can be found in time, taken measures in advance, it is to avoid pernicious Accident occurs.

Start to failure, bearing performance continuous variation from military service, form time serieses, with the performance being continually changing With reliability track.Based on suprasphere multi-class support vector machine, with improved ensemble empirical mode decomposition method and characteristic parameter heredity Optimization method, it can be estimated that bearing performance degree of degeneration；Based on minimum entropy deconvolution and autoregressive Ke Ermogeluofu-Si meter The Er Nuofu methods of inspection, abnormal phenomena when can detect that initial stage minor defect occurs in bearing；With frequency response and phase paths Nonlinear dynamic analysis are carried out etc. concept, the polytropy of bearing performance can be described.

These achievements present the mechanics of bearing performance variation and statistics rule, but are not yet related to bearing performance reliability Mutation process forecasting problem, it is impossible to the mutation process of predicted roll bearing performance reliability with it is timely find fail hidden danger, keep away Exempt from the generation of serious accident.

The content of the invention

It is an object of the invention to provide a kind of bearing vibration performance reliability mutation process detection method and device, use To solve the problems, such as the mutation process of the unpredictable rolling bearing performance reliability of existing method.

For achieving the above object, the solution of the present invention includes：

Bearing vibration performance reliability mutation process detection method, step are as follows：

1) the real-time detection bear vibration information in bearing experiment of rotation, by the vibration information be divided into one or two with On time serieses；

2) each time serieses are divided into into G subsequence；To each subsequence, by the time serieses meter to vibration information Number, in calculating subsequence, data obtain the raw information vector of variation intensity beyond the number of times of given threshold ± c；

3) the raw information vector to variation intensity, using self-service resampling method sample drawn, obtains self-service resampling Sample vector；Self-service resampling sample vector is solved according to Grey Prediction Model, variation intensity is estimated with mathematic expectaion；

4) variation intensity based on the estimation, represents bear vibration performance failure process with Poisson process, obtains bearing The reliability function of vibration performance, judges the mutation process of bear vibration performance reliability according to reliability function.

Step 1) in, definition time variable is τ；In 1 assessment cycle T, from time τ=τ_LStart timing, to time τ =τ_UTerminate timing；Take time interval Δ τ=τ_U-τ_LConstants of=the T for value very little, and represented under different time τ with subscript i Time interval, formed time interval sequence：

△ Γ=(△ τ₁,△τ₂,…,△τ_i,…,△τ_I)；I=1,2 ..., I (1)

In formula, Δ τ_iFor i-th time interval, I is time interval number, and T is assessment cycle；

X_i=(x_i(1),x_i(2),…,x_i(w),…,x_i(W))；W=1,2 ..., W (2)

In formula, X_iFor Δ τ_iInterior time interval sequence, x_iW () is X_iIn w-th data, W is X_iIn data amount check.

Step 2) in, by X_iG subsequence is divided into, each subsequence there are K data, wherein, g-th subsequence is:

X_ig=(x_ig(1), x_ig(2) ..., x_ig(k) ..., x_ig(K))；K=1,2 ..., K；G=1,2 ..., G (3)

In formula, x_igK () is X_igIn k-th data；K is X_igIn data amount check, and have

For g-th subsequence X_ig, counted by the time serieses to vibration information, calculate x_igK () is beyond the secondary of ± c Number n_ig, obtain the raw information of variation intensity：

In formula, T_gRepresent to g-th subsequence X_igCalculating cycle：

For G subsequence, the raw information vector theta of a variation intensity is built_i1：

Θ_i1={ θ_ig} (7)

Step 3) in, according to self-service resampling method, from Θ_i1Equal probability can randomly select with putting back to 1 data, extract M time, m=3,4 ..., G obtain the self-service resampling sample vector that 1 dimension is m.This process continuously repeats B steps, obtains B Individual self-service resampling sample vector, is expressed in matrix as

Y_iBootstrap={ Y_ib}_B×m；B=1,2 ..., B (8)

In formula, Y_ibIt is b-th self-service resampling sample vector, and has

Y_ib={ y_ib(j)}；J=1,2 ..., m (9)

In formula, y_ibJ () is Y_ibIn j-th self-service resampling data；

By Grey Prediction Model：If Y_ibOne-accumulate formation sequence vector be

One-accumulate formation sequence is described as with grey differential equation

In formula, c_i1And c_i2For undetermined coefficient；

If average generation sequence vector is

Z_ib={ z_ib(u) }={ 0.5 θ_ib(u)+0.5θ_ib(u-1)}；U=2,3 ..., m (12)

In initial condition θ_ib(1)=y_ib(1), under, the least square solution of grey differential equation is

In formula, coefficient c_i1And c_i2For

And have

D_i={-Z_ib,Q}^T (15)

In formula, unit vectors of the Q for dimension m-1；

By formula (13), b-th data of inverse accumulated generating can be obtained：

λ_ib=θ_1ib(m+1)-θ_1ib(m) (16)

The generation information of B variation intensity can be simulated by formula (16), with vector representation be

Θ_i={ λ_ib} (17)

Using statistical histogram method, one can be set up with regard to variation intensity by the generation information in formula (17) Probability density function：

In formula,It is the probability density function with regard to variation intensity, λ_iFor describing a stochastic variable of variation intensity；

Variation intensity is estimated with mathematic expectaion：

In formula, λ_0iFor the variation intensity estimated, Λ_iFor λ_iFeasible zone.

For i-th time interval, in an assessment cycle T, domain time variable t of setting a trap ∈ [0, T_g], based on estimation Variation intensity, the distribution law of bear vibration performance failure process represented with Poisson process：

In formula, n_iFor variation factor, the discrete of " vibration countershaft may manufacture into damage " this event frequency is represented Variable；N_iFor the frequency of event；

The cumulative distribution of bear vibration performance failure process is

The reliability function of bear vibration performance is

R_i(t)=1-P_i(t) (22)

Bearing vibration performance reliability mutation process detection means, including such as lower module：

1) the real-time detection bear vibration information in bearing experiment of rotation, by the vibration information be divided into one or two with On time serieses；

2) each time serieses are divided into into G subsequence；To each subsequence, by the time serieses meter to vibration information Number, in calculating subsequence, data obtain the raw information vector of variation intensity beyond the number of times of given threshold ± c；

3) the raw information vector to variation intensity, using self-service resampling method sample drawn, obtains self-service resampling Sample vector；Self-service resampling sample vector is solved according to Grey Prediction Model, variation intensity is estimated with mathematic expectaion；

4) variation intensity based on the estimation, represents bear vibration performance failure process with Poisson process, obtains bearing The reliability function of vibration performance, judges the mutation process of bear vibration performance reliability according to reliability function.

Module 1) in, definition time variable be τ；In 1 assessment cycle T, from time τ=τ_LStart timing, to time τ =τ_UTerminate timing；Take time interval Δ τ=τ_U-τ_LConstants of=the T for value very little, and represented under different time τ with subscript i Time interval, formed time interval sequence：

△ Γ=(△ τ₁,△τ₂,…,△τ_i,…,△τ_I)；I=1,2 ..., I (1)

In formula, Δ τ_iFor i-th time interval, I is time interval number, and T is assessment cycle；

X_i=(x_i(1),x_i(2),…,x_i(w),…,x_i(W))；W=1,2 ..., W (2)

In formula, X_iFor Δ τ_iInterior time interval sequence, x_iW () is X_iIn w-th data, W is X_iIn data amount check.

Module 2) in, by X_iG subsequence is divided into, each subsequence there are K data, wherein, g-th subsequence is:

X_ig=(x_ig(1), x_ig(2) ..., x_ig(k) ..., x_ig(K))；K=1,2 ..., K；G=1,2 ..., G (3)

In formula, x_igK () is X_igIn k-th data；K is X_igIn data amount check, and have

For g-th subsequence X_ig, counted by the time serieses to vibration information, calculate x_igK () is beyond the secondary of ± c Number n_ig, obtain the raw information of variation intensity：

In formula, T_gRepresent to g-th subsequence X_igCalculating cycle：

For G subsequence, the raw information vector theta of a variation intensity is built_i1：

Θ_i1={ θ_ig} (7)

Module 3) in, according to self-service resampling method, from Θ_i1Equal probability can randomly select with putting back to 1 data, extract M time, m=3,4 ..., G obtain the self-service resampling sample vector that 1 dimension is m.This process continuously repeats B steps, obtains B Individual self-service resampling sample vector, is expressed in matrix as

Y_iBootstrap={ Y_ib}_B×m；B=1,2 ..., B (8)

In formula, Y_ibIt is b-th self-service resampling sample vector, and has

Y_ib={ y_ib(j)}；J=1,2 ..., m (9)

In formula, y_ibJ () is Y_ibIn j-th self-service resampling data；

By Grey Prediction Model：If Y_ibOne-accumulate formation sequence vector be

One-accumulate formation sequence is described as with grey differential equation

In formula, c_i1And c_i2For undetermined coefficient；

If average generation sequence vector is

Z_ib={ z_ib(u) }={ 0.5 θ_ib(u)+0.5θ_ib(u-1)}；U=2,3 ..., m (12)

In initial condition θ_ib(1)=y_ib(1), under, the least square solution of grey differential equation is

In formula, coefficient c_i1And c_i2For

And have

D_i={-Z_ib,Q}^T (15)

In formula, unit vectors of the Q for dimension m-1；

By formula (13), b-th data of inverse accumulated generating can be obtained：

λ_ib=θ_1ib(m+1)-θ_1ib(m) (16)

The generation information of B variation intensity can be simulated by formula (16), with vector representation be

Θ_i={ λ_ib} (17)

Using statistical histogram method, one can be set up with regard to variation intensity by the generation information in formula (17) Probability density function：

In formula,It is the probability density function with regard to variation intensity, λ_iFor describing a stochastic variable of variation intensity；

Variation intensity is estimated with mathematic expectaion：

In formula, λ_0iFor the variation intensity estimated, Λ_iFor λ_iFeasible zone.

For i-th time interval, in an assessment cycle T, domain time variable t of setting a trap ∈ [0, T_g], based on estimation Variation intensity, the distribution law of bear vibration performance failure process represented with Poisson process：

In formula, n_iFor variation factor, the discrete of " vibration countershaft may manufacture into damage " this event frequency is represented Variable；N_iFor the frequency of event；

The cumulative distribution of bear vibration performance failure process is

The reliability function of bear vibration performance is

R_i(t)=1-P_i(t) (22)

The self-service Poisson method of ash is based on " vibration countershaft may manufacture into damage " this event.Inner body damage with Abrasion can cause bear vibration, and bear vibration again may the damage of aggravation inner body and abrasion.The persistent loop of this process, into For the condition of " vibration countershaft may manufacture into damage ".Due to various random disturbances, even if there be no any damage and abrasion Bearing, still can produce vibration during one's term of military service in good lubrication.Therefore, generally allow for bearing and there is vibration, but vibration can not More than threshold value.In fact, the frequency that vibration amplitude exceedes threshold value is higher, the countershaft probability for damaging of manufacturing into is bigger, reliability It is lower.According to theory of random processes, the event of this " vibration countershaft may manufacture into damage " belongs to counting process, can be with Represented with Poisson process.

According to Poisson process, based on the time serieses of vibration information, defining rolling bearing performance reliability mutation process is A counting process with variation intensity as parameter.Variation intensity refers to that vibration amplitude exceedes the frequency of threshold value, belongs to impact axle Hold the key character parameter of performance reliability mutation process.Variation intensity is as bearing performance is in different time interval variations Change.

Consider real-time estimate, the raw information for obtaining variation intensity in Short Interval is few, and forecast variation intensity Need many information.A large amount of generation information of variation intensity are simulated with self-service resampling method, is obtained with Grey Prediction Model and is become The estimated value of different intensity, it is possible to implement the mutation process prediction of bear vibration performance reliability.

Based on this, grey self-service principle is incorporated Poisson process by time serieses of the present invention based on vibration information, proposes ash certainly Poisson method is helped, with the mutation process of predicted roll bearing performance reliability.

Description of the drawings

Fig. 1 bear vibration time serieses X₁(d₁=0mm)；

Fig. 2 bear vibration time serieses X₂(d₂=0.1778mm)；

Fig. 3 bear vibration time serieses X₃(d₃=0.5334mm)；

Fig. 4 bear vibration time serieses X₄(d₄=0.7112mm)；

Distribution law (the d of Fig. 5 bear vibration performance variation processes₂=0.1778mm)；

Distribution law (the d of Fig. 6 bear vibration performance variation processes₃=0.5334mm)；

Distribution law (the d of Fig. 7 bear vibration performance variation processes₄=0.7112mm)；

Fig. 8 reliability predictions and inspection；

Reliability prediction value sequence under Fig. 9 difference threshold values.

Specific embodiment

The present invention will be further described in detail below in conjunction with the accompanying drawings.

Tested according to prior failure first, obtain the threshold value ± c of bear vibration performance., the threshold value is obtained by experiment , different with sensor according to the damage location chosen, corresponding threshold value is different.

Definition time variable is τ.In 1 assessment cycle T, from different time τ=τ_LStart timing, to time τ=τ_U Terminate timing.Take time interval Δ τ=τ_U-τ_L=T for value very little constant, and with subscript i represent under different time τ when Between it is interval, form a Short Interval sequence：

△ Γ=(△ τ₁,△τ₂,…,△τ_i,…,△τ_I)；I=1,2 ..., I (1)

In formula, Δ τ_iFor i-th time interval, I is time interval number, and T is assessment cycle.

Assume, in i-th time interval, to obtain the time of bear vibration information during one's term of military service by measuring system Sequence of intervals X_i：

X_i=(x_i(1),x_i(2),…,x_i(w),…,x_i(W))；W=1,2 ..., W (2)

In formula, x_iW () is X_iIn w-th data, W is X_iIn data amount check.

For impact of the random noise attenuation to predicting the outcome, by X_iG subsequence is divided into, each subsequence has K Data, wherein, g-th subsequence is

X_ig=(x_ig(1), x_ig(2) ..., x_ig(k) ..., x_ig(K))；K=1,2 ..., K；G=1,2 ..., G (3)

In formula, x_igK () is X_igIn k-th data；K is X_igIn data amount check, and have

For g-th subsequence X_ig, counted by the time serieses to vibration information, calculate x_igK () is beyond the secondary of ± c Number n_ig, obtain the raw information of variation intensity：

In formula, T_gRepresent to g-th subsequence X_igCalculating cycle：

For G subsequence, the raw information vector theta of a variation intensity can be built_i1：

Θ_i1={ θ_ig} (7)

Based on Θ_i1, the estimated value of variation intensity can be extracted, is that prediction bear vibration performance reliability mutation process is established Determine parameter basis.

Using grey self-service principle, the estimated value of variation intensity is obtained.

According to self-service resampling method, from Θ_i1Equal probability can randomly select with putting back to 1 data, extract m time (m=3, 4 ..., G), obtain the self-service resampling sample vector that 1 dimension is m.This process continuously repeats B steps and (generally can use B= 1000～500000), B self-service resampling sample vector is obtained, is expressed in matrix as

Y_iBootstrap={ Y_ib}_B×m；B=1,2 ..., B (8)

In formula, Y_ibIt is b-th self-service resampling sample vector, and has

Y_ib={ y_ib(j)}；J=1,2 ..., m (9)

In formula, y_ibJ () is Y_ibIn j-th self-service resampling data.

By Grey Prediction Model, if Y_ibOne-accumulate formation sequence vector be

One-accumulate formation sequence can be described as with grey differential equation

In formula, u can regard a continuous variable, c as_i1And c_i2For undetermined coefficient.

If average generation sequence vector is

Z_ib={ z_ib(u) }={ 0.5 θ_ib(u)+0.5θ_ib(u-1)}；U=2,3 ..., m (12)

In initial condition θ_ib(1)=y_ib(1), under, the least square solution of grey differential equation is

In formula, coefficient c_i1And c_i2For

And have

D_i={-Z_ib,Q}^T (15)

In formula, unit vectors of the Q for dimension m-1.

By formula (13), b-th data of inverse accumulated generating can be obtained：

λ_ib=θ_1ib(m+1)-θ_1ib(m) (16)

The generation information of B variation intensity can be simulated by formula (16), with vector representation be

Θ_i={ λ_ib} (17)

Using statistical histogram method, one can be set up with regard to variation intensity by the generation information in formula (17) Probability density function：

In formula,It is the probability density function with regard to variation intensity, λ_iFor describing a stochastic variable of variation intensity.

Variation intensity is estimated with mathematic expectaion：

In formula, λ_0iFor the variation intensity estimated, Λ_iFor λ_iFeasible zone.

For i-th time interval, in an assessment cycle T, domain time variable t of setting a trap ∈ [0, T_g], based on estimation Variation intensity, the distribution law of bear vibration performance failure process represented with Poisson process：

In formula, n_iFor variation factor, the discrete of " vibration countershaft may manufacture into damage " this event frequency is represented Variable；N_iFor the frequency of event.

The cumulative distribution of bear vibration performance failure process is

The reliability function of bear vibration performance is

R_i(t)=1-P_i(t) (22)

Different from traditional bearing life reliability, the t in formula (22) is not life variance, but is based on time variable τ Local time variable.With the operation of bearing, Δ τ can be extracted_iInterior R_iT (), then obtains with regard to R_iThe function of (t) Sequence：

R=(R₁(t),R₂(t),…,R_i(t),…,R_I(t)) (23)

In formula, R is reliability function sequence, R_iT () is the reliability function in i-th time interval.

Make t=T_g, calculate R_iT the value of (), then obtains the prediction value sequence of reliability：

R=(r₁,r₂,…,r_i,…,r_I) (24)

In formula

r_i=R_i(T) (25)

In formula, r_iFor the predictive value of i-th time interval inner bearing vibration performance reliability.

According to the situation of change of the variable τ over time of reliability prediction value sequence in formula (25), can be with real-time estimate bearing The mutation process of vibration performance reliability.In bearing during one's term of military service, corresponding strick precaution can be taken in time to arrange according to alert level Apply, to avoid serious accident from occurring.

According to statistical Little Probability Event Princiole, take small probability value and be respectively 0.01,0.05 and 0.10 etc., it is corresponding can 99%, 95% and 90% etc. is respectively by property value, is 1～6 totally 6 alert levels by bear vibration behavioral definition：

If r_i∈ [99%, 100%], then bear vibration performance is 1 grade, there is a possibility that potential safety hazard very little；

If r_i[95%, 99%), then bear vibration performance is 2 grades to ∈, there is a possibility that potential safety hazard is little；

If r_i[90%, 95%), then bear vibration performance is 3 grades to ∈, there is a possibility that potential safety hazard is smaller；

If r_i[85%, 90%), then bear vibration performance is 4 grades to ∈, there is potential safety hazard；

If r_i[75%, 85%), then bear vibration performance is 5 grades to ∈, there is big potential safety hazard；

If r_i<75%, then bear vibration performance is 6 grades, there is very big potential safety hazard.

Above method is applied to channel surface abrasion causes bear vibration acceleration to morph, will the inventive method In variation specific to bearing inner race raceway groove damage, see following example：

Experimental data, should from the bearing data center website of U.S. Case Western Reserve University Center possesses a special rolling bearing fault simulated experiment platform.Laboratory table is by motor, torque sensor/decoder and work( Rate tester etc. is constituted.SKF6205 bearings to be detected the gyroaxis of motor.Bearing is measured with acceleration transducer Acceleration of vibration, unit are V.Bearing rotating speed is 1797r/min, and sample frequency is 12kHz, bearing inner race raceway groove lesion diameter d Respectively d₁=0mm, d₂=0.1778mm, d₃=0.5334mm and d₄=0.7112mm.

The time serieses of the bear vibration information for being obtained are as shown in Figure 1 to 4.As can be seen that lesion diameter is bigger, axle Hold vibration it is more violent, failure probability is bigger.Therefore can be assessed in bearing by the variation of analysis vibration performance reliability Portion's part injury and abrasion condition.

By X_i4 subsequences are divided into, forecast model are set up with front 3 subsequences (w=0～1200), implement reliability Prediction；Prediction effect is verified with last 1 subsequence (w=1200～1600), and 4 kinds of lesion diameters are modeled as into bearing experience Amount of damage produced by the operation of 4 time intervals (i=1,2,3,4).

When forecast model is set up, G=3, W=1200, K=400, B=400000, T is taken_g=0.033s, I=4, c= 0.5V。

The raw information vector theta of different amount of damage lower variation intensity is obtained by formula (1)～formula (7)_i1, by formula (8)～formula (19) obtain estimated value λ of different amount of damage lower variation intensity_i0, the results are shown in Table 1.As shown in Table 1, under given amount of damage, Raw information θ of bear vibration each subsequence variation intensity_ig(g=1,2,3) variant, this is that random noise causes.It is logical Cross grey bootstrap filter forecasting to go out, amount of damage is bigger, variation intensity λ_i0It is bigger.This shows, with the growth of run time τ, axle Inner body is held in each time interval Δ τ_iDamage phenomenon become increasingly severe, cause variation intensity constantly to increase, most The continuous worsening of vibration performance is shown as eventually, buries failure hidden danger.

Distribution law p of bear vibration performance failure process is obtained by formula (20)_i(n_i, t), as shown in Fig. 5～Fig. 7.

The raw information and estimated value of 1 variation intensity of table

As can be seen that with the growth of run time τ, in different run time interval Δ τ_iIt is interior, p_i(n_i, shape t) It is each with orientation variant.This is different variation intensities λ_i0Cause.In given Δ τ_iInterior (d_iValue determines), with n_iWith t's Increase, p_i(n_i, peak value t) reduces and width increase, the uncertain rising that the event that represents occurs.

In calculating cycle T_gAt the end of, event frequency reaches capacity, and the failure probability of accumulation is maximum.

Can be obtained under different amount of damage by formula (25), bearing runs the reliability prediction value sequence of a cycle, knot Fruit is as shown in Figure 8.As can be seen that in the running for experiencing 4 time intervals, with the increase of amount of damage, bear vibration Performance reliability is presented non-linear downward trend.Concrete rule can be divided into 3 stages：When amount of damage gradually increases from 0, Reliability decrease is very fast, belongs to for the 1st stage；When amount of damage continues to increase, reliability decrease is slow and has micro fluctuation, belongs to 2nd stage；When amount of damage exceedes certain value, reliability decrease quickly, belonged to for the 3rd stage.Reliability variation law forms one The chaise longue shape curve of individual non-linear decline.This is that variation intensity change is caused, and disclosing bear vibration performance time sequence can By the inherent Variation mechanism of property.

In order to verify the correctness of reliability prediction value variation law, by last 1 subsequence (w=in Fig. 1～Fig. 4 1200～1600) as checking sequence, the corresponding reliability value of calculating, and result is regarded as the test value of reliability, represent In fig. 8.Be not difficult to find out, reliability prediction value is identical with the Changing Pattern of test value, concordance very well, error therebetween Very little, maximum absolute error value are 0.068, and maximum relative error value is 14.5%.

Fig. 9 is the reliability prediction value sequence under different threshold values.As can be seen that threshold value is less, reliability is lower；Conversely, Reliability is higher.Therefore, in engineering practice, according to requirement of the concrete system to bearing vibration performance, failure is carried out in advance real Test, obtain threshold value.During service, real-time detection is carried out to bearing vibration information and obtains reliability prediction value, can be timely It was found that failure hidden danger, it is to avoid serious accident occurs.

In above example, " vibration countershaft may manufacture into damage " this event is specifically limited as bearing inner race raceway groove Damage, represented by lesion diameter.In order to express lesion diameter, the present invention adopts electric parameters magnitude of voltage.In fact, based on biography Sensor is different from detection circuit, it is also possible to represented with other physical quantitys.

With regard to assessment cycle T, can be continuous, or discontinuous between each assessment cycle.

It is specific embodiment to be given above, but the present invention is not limited to described embodiment.The present invention's Basic ideas are such scheme, and for those of ordinary skill in the art, various modifications are designed in teaching of the invention Model, formula, parameter creative work need not be spent.Without departing from the principles and spirit of the present invention to reality Change, modification, replacement and the modification that the mode of applying is carried out is still fallen within protection scope of the present invention.

Claims (10)

1. bearing vibration performance reliability mutation process detection method, it is characterised in that step is as follows：

1) vibration information is divided into one or more by the real-time detection bear vibration information in bearing experiment of rotation Time serieses；

2) each time serieses are divided into into G subsequence；To each subsequence, counted by the time serieses to vibration information, In calculating subsequence, data obtain the raw information vector of variation intensity beyond the number of times of given threshold ± c；

3) the raw information vector to variation intensity, using self-service resampling method sample drawn, obtains self-service resampling sample Vector；Self-service resampling sample vector is solved according to Grey Prediction Model, variation intensity is estimated with mathematic expectaion；

4) variation intensity based on the estimation, represents bear vibration performance failure process with Poisson process, obtains bear vibration The reliability function of performance, judges the mutation process of bear vibration performance reliability according to reliability function.

2. bearing vibration performance reliability mutation process detection method according to claim 1, it is characterised in that step It is rapid 1) in, definition time variable be τ；In 1 assessment cycle T, from time τ=τ_LStart timing, to time τ=τ_UTerminate meter When；Take time interval Δ τ=τ_U-τ_LConstants of=the T for value very little, and the time interval under different time τ is represented with subscript i, Form time interval sequence：

△ Γ=(△ τ₁,△τ₂,…,△τ_i,…,△τ_I)；I=1,2 ..., I (1)

In formula, Δ τ_iFor i-th time interval, I is time interval number, and T is assessment cycle；

X_i=(x_i(1),x_i(2),…,x_i(w),…,x_i(W))；W=1,2 ..., W (2)

In formula, X_iFor Δ τ_iInterior time interval sequence, x_iW () is X_iIn w-th data, W is X_iIn data amount check.

3. bearing vibration performance reliability mutation process detection method according to claim 2, it is characterised in that Step 2) in, by X_iG subsequence is divided into, each subsequence there are K data, wherein, g-th subsequence is:

X_ig=(x_ig(1),x_ig(2),…,x_ig(k),…,x_ig(K))；K=1,2 ..., K；G=1,2 ..., G (3)

In formula, x_igK () is X_igIn k-th data；K is X_igIn data amount check, and have

For g-th subsequence X_ig, counted by the time serieses to vibration information, calculate x_igThe frequency n of (k) beyond ± c_ig, Obtain the raw information of variation intensity：

In formula, T_gRepresent to g-th subsequence X_igCalculating cycle：

For G subsequence, the raw information vector theta of a variation intensity is built_i1：

Θ_i1={ θ_ig} (7) 。

4. bearing vibration performance reliability mutation process detection method according to claim 3, it is characterised in that step It is rapid 3) in, according to self-service resampling method, from Θ_i1Equal probability can randomly select with putting back to 1 data, extract m time, m=3, 4 ..., G, obtain the self-service resampling sample vector that 1 dimension is m, and this process continuously repeats B steps, obtain that B is self-service to be taken out again Sample sample vector, is expressed in matrix as

Y_iBootstrap={ Y_ib}_B×m；B=1,2 ..., B (8)

In formula, Y_ibIt is b-th self-service resampling sample vector, and has

Y_ib={ y_ib(j)}；J=1,2 ..., m (9)

In formula, y_ibJ () is Y_ibIn j-th self-service resampling data；

By Grey Prediction Model：If Y_ibOne-accumulate formation sequence vector be

One-accumulate formation sequence is described as with grey differential equation

In formula, c_i1And c_i2For undetermined coefficient；

If average generation sequence vector is

Z_ib={ z_ib(u) }={ 0.5 θ_ib(u)+0.5θ_ib(u-1)}；U=2,3 ..., m (12)

In initial condition θ_ib(1)=y_ib(1), under, the least square solution of grey differential equation is

In formula, coefficient c_i1And c_i2For

And have

D_i={-Z_ib,Q}^T (15)

In formula, unit vectors of the Q for dimension m-1；

By formula (13), b-th data of inverse accumulated generating can be obtained：

λ_ib=θ_1ib(m+1)-θ_1ib(m) (16)

The generation information of B variation intensity can be simulated by formula (16), with vector representation be

Θ_i={ λ_ib} (17)

Using statistical histogram method, a probability with regard to variation intensity can be set up by the generation information in formula (17) Density function：

In formula,It is the probability density function with regard to variation intensity, λ_iFor describing a stochastic variable of variation intensity；

Variation intensity is estimated with mathematic expectaion：

In formula, λ_0iFor the variation intensity estimated, Λ_iFor λ_iFeasible zone.

5. bearing vibration performance reliability mutation process detection method according to claim 4, it is characterised in that right In i-th time interval, in an assessment cycle T, domain time variable t of setting a trap ∈ [0, T_g], based on the variation intensity estimated, The distribution law of bear vibration performance failure process is represented with Poisson process：

In formula, n_iFor variation factor, the discrete variable of " vibration countershaft may manufacture into damage " this event frequency is represented； N_iFor the frequency of event；

The cumulative distribution of bear vibration performance failure process is

The reliability function of bear vibration performance is

R_i(t)=1-P_i(t) (22) 。

6. bearing vibration performance reliability mutation process detection means, it is characterised in that include such as lower module：

1) vibration information is divided into one or more by the real-time detection bear vibration information in bearing experiment of rotation Time serieses；

2) each time serieses are divided into into G subsequence；To each subsequence, counted by the time serieses to vibration information, In calculating subsequence, data obtain the raw information vector of variation intensity beyond the number of times of given threshold ± c；

3) the raw information vector to variation intensity, using self-service resampling method sample drawn, obtains self-service resampling sample Vector；Self-service resampling sample vector is solved according to Grey Prediction Model, variation intensity is estimated with mathematic expectaion；

4) variation intensity based on the estimation, represents bear vibration performance failure process with Poisson process, obtains bear vibration The reliability function of performance, judges the mutation process of bear vibration performance reliability according to reliability function.

7. bearing vibration performance reliability mutation process detection means according to claim 6, it is characterised in that mould Block 1) in, definition time variable be τ；In 1 assessment cycle T, from time τ=τ_LStart timing, to time τ=τ_UTerminate meter When；Take time interval Δ τ=τ_U-τ_LConstants of=the T for value very little, and the time interval under different time τ is represented with subscript i, Form time interval sequence：

△ Γ=(△ τ₁,△τ₂,…,△τ_i,…,△τ_I)；I=1,2 ..., I (1)

In formula, Δ τ_iFor i-th time interval, I is time interval number, and T is assessment cycle；

X_i=(x_i(1),x_i(2),…,x_i(w),…,x_i(W))；W=1,2 ..., W (2)

In formula, X_iFor Δ τ_iInterior time interval sequence, x_iW () is X_iIn w-th data, W is X_iIn data amount check.

8. bearing vibration performance reliability mutation process detection means according to claim 7, it is characterised in that mould Block 2) in, by X_iG subsequence is divided into, each subsequence there are K data, wherein, g-th subsequence is:

X_ig=(x_ig(1),x_ig(2),…,x_ig(k),…,x_ig(K))；K=1,2 ..., K；G=1,2 ..., G (3)

In formula, x_igK () is X_igIn k-th data；K is X_igIn data amount check, and have

For g-th subsequence X_ig, counted by the time serieses to vibration information, calculate x_igThe frequency n of (k) beyond ± c_ig, Obtain the raw information of variation intensity：

In formula, T_gRepresent to g-th subsequence X_igCalculating cycle：

For G subsequence, the raw information vector theta of a variation intensity is built_i1：

Θ_i1={ θ_ig} (7) 。

9. bearing vibration performance reliability mutation process detection means according to claim 8, it is characterised in that mould Block 3) in, according to self-service resampling method, from Θ_i1Equal probability can randomly select with putting back to 1 data, extract m time, m=3, 4 ..., G, obtain the self-service resampling sample vector that 1 dimension is m, and this process continuously repeats B steps, obtain that B is self-service to be taken out again Sample sample vector, is expressed in matrix as

Y_iBootstrap={ Y_ib}_B×m；B=1,2 ..., B (8)

In formula, Y_ibIt is b-th self-service resampling sample vector, and has

Y_ib={ y_ib(j)}；J=1,2 ..., m (9)

In formula, y_ibJ () is Y_ibIn_jIndividual self-service resampling data；

By Grey Prediction Model：If Y_ibOne-accumulate formation sequence vector be

One-accumulate formation sequence is described as with grey differential equation

In formula, c_i1And c_i2For undetermined coefficient；

If average generation sequence vector is

Z_ib={ z_ib(u) }={ 0.5 θ_ib(u)+0.5θ_ib(u-1)}；U=2,3 ..., m (12)

In initial condition θ_ib(1)=y_ib(1), under, the least square solution of grey differential equation is

In formula, coefficient c_i1And c_i2For

And have

D_i={-Z_ib,Q}^T (15)

In formula, unit vectors of the Q for dimension m-1；

By formula (13), b-th data of inverse accumulated generating can be obtained：

λ_ib=θ_1ib(m+1)-θ_1ib(m) (16)

The generation information of B variation intensity can be simulated by formula (16), with vector representation be

Θ_i={ λ_ib} (17)

Using statistical histogram method, a probability with regard to variation intensity can be set up by the generation information in formula (17) Density function：

In formula,It is the probability density function with regard to variation intensity, λ_iFor describing a stochastic variable of variation intensity；

Variation intensity is estimated with mathematic expectaion：

In formula, λ_0iFor the variation intensity estimated, Λ_iFor λ_iFeasible zone.

10. bearing vibration performance reliability mutation process detection means according to claim 9, it is characterised in that For i-th time interval, in an assessment cycle T, domain time variable t of setting a trap ∈ [0, T_g], it is strong based on the variation estimated Degree, the distribution law of bear vibration performance failure process are represented with Poisson process：

In formula, n_iFor variation factor, the discrete variable of " vibration countershaft may manufacture into damage " this event frequency is represented； N_iFor the frequency of event；

The cumulative distribution of bear vibration performance failure process is

The reliability function of bear vibration performance is

R_i(t)=1-P_i(t) (22) 。