CN111444988A - Rolling bearing fault diagnosis system - Google Patents

Rolling bearing fault diagnosis system Download PDF

Info

Publication number
CN111444988A
CN111444988A CN202010389878.4A CN202010389878A CN111444988A CN 111444988 A CN111444988 A CN 111444988A CN 202010389878 A CN202010389878 A CN 202010389878A CN 111444988 A CN111444988 A CN 111444988A
Authority
CN
China
Prior art keywords
signal
entropy
decomposition
fault diagnosis
rolling bearing
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010389878.4A
Other languages
Chinese (zh)
Other versions
CN111444988B (en
Inventor
牛国成
胡冬梅
周振雄
孙继元
孟祥�
刘柏生
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beihua University
Original Assignee
Beihua University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beihua University filed Critical Beihua University
Priority to CN202010389878.4A priority Critical patent/CN111444988B/en
Publication of CN111444988A publication Critical patent/CN111444988A/en
Application granted granted Critical
Publication of CN111444988B publication Critical patent/CN111444988B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00—Testing of machine parts
    • G01M13/04—Bearings
    • G01M13/045—Acoustic or vibration analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Probability & Statistics with Applications (AREA)
  • Acoustics & Sound (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
  • Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)

Abstract

The invention discloses a rolling bearing fault diagnosis system. Aiming at the problem of fault diagnosis of the rolling bearing, the dual-filtering signal feature extraction method combines a permutation entropy-assisted combined modal decomposition algorithm with a singular value decomposition algorithm. The combined modal decomposition algorithm not only realizes the primary filtering of signals, but also overcomes the modal aliasing phenomenon and improves the calculation instantaneity. Singular value decomposition secondary filtering further accurately extracts the characteristic frequency of the bearing running state; the fault state of the bearing can be represented from multiple angles by using the multi-scale spreading entropy, and the fault diagnosis efficiency of the rolling bearing is improved to a great extent by combining the multi-scale spreading entropy with the Gaussian mixture continuous hidden Markov fault diagnosis model.

Description

Rolling bearing fault diagnosis system
Technical Field
The invention relates to the technical field of machinery, in particular to a rolling bearing fault diagnosis system.
Background
Rolling bearings are "joints" connecting rotating parts and fixed parts in mechanical equipment, and their operating states determine the operating conditions of the mechanical equipment and thus the production line. The failure rate of the rotary machine due to the rolling bearing is about 30%. The current technologies for diagnosing the faults of the rolling bearing mainly comprise: vibration diagnosis technology, acoustic diagnosis technology, oil film resistance diagnosis technology, temperature diagnosis technology, ferrograph diagnosis technology and the like, more than 80% of the existing research documents about rolling bearing fault diagnosis adopt a vibration signal analysis method.
The invention integrates empirical mode decomposition and complete set empirical mode decomposition to form a combined modal component decomposition feature extraction technology, and combines a singular value decomposition technology to realize secondary noise reduction and feature extraction of the rolling bearing. And recognizing and classifying the running state of the bearing by adopting a mixed Gaussian continuous hidden Markov model based on multi-scale diffusion entropy. The invention can make the failure diagnosis rate of the rolling bearing reach 100 percent, and improve about 10 percent compared with the prior research result. The running state of the rolling bearing is accurately analyzed through the fault diagnosis method, scientific operation and maintenance of bearing equipment and reasonable production arrangement are realized, the service life of the equipment is prolonged, the cost is reduced, and the production efficiency is improved.
Disclosure of Invention
Aiming at the defects of the related technology of the fault diagnosis of the existing rolling bearing, the invention provides a fault diagnosis system of the rolling bearing, which can accurately identify the running state of the rolling bearing and improve the fault diagnosis efficiency of the rolling bearing.
The invention is realized by the following technical scheme:
a rolling bearing fault diagnosis system characterized in that:
the first step is as follows: the combined modal decomposition primary filtering of the fault signals of the rolling bearing comprises the following steps:
1) selecting vibration signals of a vibration sensor of the bearing and carrying out signal grouping, wherein the signal grouping is to divide time continuous signals in a time period and calculate the arrangement entropy of each group of original vibration signals;
a. if the permutation entropy is less than 0.618, selecting intrinsic empirical mode decomposition to carry out original signal decomposition, decomposing to obtain intrinsic modal components and residual amount, and entering b;
if the permutation entropy is larger than 0.618, carrying out complete set empirical mode decomposition on the self-adaptive noise to obtain a natural modal component and a residual amount, and entering b;
b. the number of the extreme value points of the residual quantity is judged,
if the number of the extreme points is more than or equal to 3, repeating the operation a to obtain the next inherent modal component and residual quantity;
if the number of the extreme values is less than 3, adding all the inherent modal components obtained by decomposition to obtain the inherent modal component after the initial filtering of the original signal;
2) selecting inherent modal components of an original signal after primary filtering, calculating a correlation coefficient and a kurtosis value, and selecting the inherent modal components of which the correlation coefficient is greater than 0.1 and the kurtosis is greater than 3; sequentially superposing the intrinsic modal components according to the kurtosis values from large to small, and calculating the kurtosis value of a superposed signal once every superposition; finally, selecting the superposed signal with the maximum kurtosis value as a final reconstruction signal;
the second step is that: the second filtering of the signal by singular value decomposition is carried out as follows:
constructing the vibration signal with the length of N and after filtering output in the first step into an m-row N-column Hankel matrix, establishing rules that the row number m is N/2(N is an even number) or (N +1)/2(N is an odd number) and the column number N is N +1-m, m is more than or equal to 2, N is more than 2 and m is less than N, performing singular value decomposition on the Hankel matrix, determining the effective order of singular values by using a singular value difference spectrum maximum method to obtain a new singular value matrix S' reflecting the signal characteristics, and performing singular value decomposition inverse operation according to the matrix to obtain a secondary reconstruction signal;
the third step: and extracting an input feature vector of fault diagnosis, wherein the process is as follows:
after the vibration state signal of the rolling bearing is subjected to combined modal decomposition filtering and singular value decomposition filtering, performing characteristic representation by using a multi-scale diffusion entropy; factors that determine the effect on the multiscale spread entropy are the embedding dimension m, the delay time d, and the class c, where m is 2, c is 6, d is 1, and the maximum time is takenInterval taumaxCalculating a multi-scale spreading entropy value of the secondary reconstruction signal as an input feature vector of fault diagnosis, wherein the multi-scale spreading entropy value is 20;
the fourth step: establishing a Gaussian mixture continuous hidden Markov rolling bearing fault diagnosis model:
processing labeled bearing fault historical data through a first step, a second step and a third step, calculating to obtain multi-scale dispersion entropy of a reconstructed signal as a training sample, and calculating model parameters lambda (pi, A and B) of a Gaussian mixture continuous hidden Markov model, wherein A is { a ═ BijIs the state transition probability distribution, pi ═ piiB, fitting the probability distribution of the observed values by using Gaussian mixture distribution to establish a rolling bearing fault diagnosis model;
the fifth step: and processing the collected vibration signal data of the operating bearing through the first step, the second step and the third step, calculating to obtain the multi-scale dispersion entropy of the reconstructed signal, inputting the multi-scale dispersion entropy into the model established in the fourth step, and determining the operating state of the vibration signal by adopting a maximum likelihood probability comparison method.
Compared with the prior art, the invention has the beneficial effects that: aiming at the problem of fault diagnosis of the rolling bearing, the dual-filtering signal feature extraction method combines a permutation entropy-assisted combined modal decomposition algorithm with a singular value decomposition algorithm. The combined modal decomposition algorithm not only realizes the primary filtering of signals, but also overcomes the modal aliasing phenomenon and improves the calculation instantaneity. Singular value decomposition secondary filtering further accurately extracts the characteristic frequency of the bearing running state; the fault state of the bearing can be represented from multiple angles by using the multi-scale spreading entropy, and the fault diagnosis efficiency of the rolling bearing is improved to a great extent by combining the multi-scale spreading entropy with the Gaussian mixture continuous hidden Markov fault diagnosis model.
Drawings
FIG. 1 is an overall workflow diagram of the present invention.
Fig. 2 is a flow chart of the intrinsic empirical mode decomposition.
FIG. 3 is a flow chart of a complete set empirical mode decomposition algorithm for adaptive noise.
FIG. 4 is a flow chart of a Gaussian mixture continuous hidden Markov rolling bearing fault diagnosis model.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1:
the first step is as follows: the combined modal decomposition primary filtering of the fault signals of the rolling bearing comprises the following implementation processes:
1. selecting vibration signal data of the ground rolling bearing, calculating the permutation entropy of the selected signals, and if the permutation entropy is less than 0.618, selecting the intrinsic empirical mode decomposition to decompose the original signals, wherein the decomposition process is shown in fig. 2:
1) constructing the upper envelope v of the vibration signal x (t)1(t) and the lower envelope v2(t) fitting an upper envelope curve and a lower envelope curve containing minimum points of all local maximum values x (t) by using a cubic spline interpolation method to enable all the minimum points to be positioned in the envelope range of the upper envelope curve and the lower envelope curve;
2) taking the average value m of the upper envelope and the lower envelope1(t), namely:
Figure BDA0002485374940000041
3) decompose x (t) to obtain the intrinsic empirical mode components,
(1) the decomposition yields a first natural modal component imf1(t)，
Taking original signal x (t) and envelope mean m1The difference of (t) is h1(t), namely:
h1(t)＝x(t)-m1(t),
condition 1: the difference between the total number of all extreme values of the original decomposed signal sequence and the sum of the total number of zero-crossing points is less than or equal to 1;
condition 2: at any time, the average value of the upper envelope and the lower envelope formed by the local maximum value and the local minimum value of the decomposed signal is zero, namely, the requirement of the symmetry of the time axes of the upper envelope and the lower envelope is met.
If h is1(t) satisfying both the requirements of condition 1 and condition 2, and subjecting h1(t) imf viewed as x (t)1(t), namely:
imf1(t)＝h1(t),
if the requirements of the condition 1 and the condition 2 are not met simultaneously, h is added1(t) regarding the new original data x (t), repeating the steps 1) -3), and judging h again1(t) mean value m of upper and lower envelope lines11(t) if the difference satisfies the normal mode component definition condition, imf1(t)＝h1(t)-m11(t)＝h11(t), if not, x (t) h11(t) circularly executing the steps from 1) to 3) until the inherent modal component generated by the k-th decomposition meets the requirement of the assumed condition, and enabling h1k＝h1(k-1)-m1kThen h is1k(t) is imf1(t), namely:
imf1(t)＝h1k(t),
(2) the i-th inherent modal component imf is obtained by decompositioni(t),
Imf will be mixed1The first residual component of the original signal, denoted as r, is subtracted from x (t)1(t)，
r1(t)＝x(t)-imf1(t),
(3) Calculating r1(t) entropy of permutation, if greater than 0.618, go to 2 steps, if less than 0.618, for r1(t) repeating steps 1) -3) above to obtain a second natural modal component imf2All residual components ri(t) the arrangement entropy is less than 0.618, the steps 1) -3) can be repeated, and the other inherent modal components of each order can be obtained, and all the residual component sequences are represented as ri(t) (i ═ 1, 2.., N), when r is greater than ri(t) (i is more than or equal to 1 and less than or equal to N) is a monotonous function, namely the original signal decomposition is finished when the number of extreme points is less than 3.
4) The original signal can be represented as a residual quantity rn(t) and natural modes of respective ordersThe sum of the components, i.e.:
Figure BDA0002485374940000051
2. if the permutation entropy is larger than 0.618, a complete set empirical mode decomposition of the adaptive noise is performed, as shown in fig. 3:
1) complete set empirical mode decomposition of adaptive noise adds a noise signal x to the original vibration signali(t)＝x(t)+ini(t)，xi(t) is the vibration signal after adding noise, ni(t) is a white noise signal,ii is (1,2, … n) the noise amplitude. Decomposing by using an empirical mode decomposition algorithm, and averaging the first-order natural mode components of each white noise signal to obtain an average value
Figure BDA0002485374940000061
For the first natural modal component:
Figure BDA0002485374940000062
2) the residual amount of the first order natural mode component is denoted as r1(t) the calculation formula is:
Figure BDA0002485374940000063
3) calculating r1(t) if the permutation entropy is less than 0.618, the step 1 is carried out, if the permutation entropy is more than 0.618, i (i is 1,2, … N) times of experiments are carried out, and the signal is represented as r1(t)+1E1(ni(t))，r1(t) residual amounts of first order natural modal components,1is the first white noise amplitude, E1(ni(t)) is a mathematical expectation of white noise. And to the signal r1(t)+1E1(ni(t)) decomposing until the first order natural modal component is obtained, and further completing second order natural modal component calculation:
Figure BDA0002485374940000064
4) after each decomposition, whether the permutation entropy of the residual components is greater than 0.618 is needed to select to enter 1-step intrinsic empirical mode decomposition or 2-step complete set empirical mode decomposition of adaptive noise, and when the permutation entropy of all the residual components is greater than 0.618, when the permutation entropy of the intrinsic modal components is 2, … M, the jth margin and the jth + 1-th intrinsic modal component are calculated according to the calculation method in the step 3):
Figure BDA0002485374940000065
5) adding 1 to j, returning to step 4), when j decomposes the residue rjAnd (t) stopping the decomposition operation when the number of the extreme points is less than 3. Then, M-order natural modal components are shared, and the residual quantity is:
Figure BDA0002485374940000071
then x (t) may be represented by a complete set of empirical mode decomposition signals of adaptive noise as:
Figure BDA0002485374940000072
3. selecting inherent modal components of an original signal after primary filtering, calculating a correlation coefficient and a kurtosis value, and selecting the inherent modal components of which the correlation coefficient is greater than 0.1 and the kurtosis is greater than 3; sequentially superposing the intrinsic modal components according to the kurtosis values from large to small, and calculating the kurtosis value of a superposed signal once every superposition; and finally, selecting the superposed signal with the maximum kurtosis value as a final reconstruction signal, wherein the kurtosis value and a related coefficient are obtained by the following steps:
1) the correlation coefficient selects a parameter. Two time series xiAnd yiCorrelation coefficient ρxyThe mathematical description is:
Figure BDA0002485374940000073
where rhoxy∈[-1,1]，|ρxyThe smaller | is, the closer the reconstructed signal is to the original signal is, the better the noise reduction effect is, and the | ρ is takenxyI is greater than 0.1.
2) Kurtosis values select parameters. The kurtosis is an index parameter of characteristic distribution of a time domain signal x (t), and the mathematical expression of the kurtosis is as follows:
Figure BDA0002485374940000074
in the formula, xi、
Figure BDA0002485374940000075
σ is the mean, standard deviation, respectively, of the ith value of the signal x (t) of length N. When the kurtosis value is greater than 3, the impact component is more obvious when the kurtosis value is larger.
The second step is that: the second filtering of the signal by singular value decomposition is carried out as follows:
selecting an inherent modal component signal X ═ X (1), X (2), …, X (N) after primary filtering of an original signal, and constructing an m-row and n-column Hankel matrix by using signals in order to decompose the signals into a trunk signal reflecting a signal main body and a series of detail signals reflecting local characteristics.
Figure BDA0002485374940000081
Wherein m is more than or equal to 2, N is more than 2, m is less than N, and m + N-1 is equal to N. When N is an even number, taking m as N/2 and N as N/(2+ 1); when N is an odd number, m ═ N +1)/2 is taken. Subjecting the matrix A to singular value decomposition, i.e.
A＝u1λ1v1 T+u2λ2v2 T+…+ukλkvk T+…+uqλqvq T,k＝1,2,…,q
The first k singular values contain the main information of the signal, while the noise is distributed over the latter singular values. Reconstructing the signal therefore requires selecting the k singular values in front of the matrix a and zeroing the q-k singular values thereafter.
The useful components are selected by a singular value difference spectrum method in singular value decomposition signal denoising, and the calculation formula is as follows: sigmaiThe decomposed ith singular value i is (1,2, … q) and is normalized singular value
Figure BDA0002485374940000082
The ith singular value difference spectrum is b (i),
Figure BDA0002485374940000083
ratio of first two singular values σ1/σ2And performing singular value decomposition reconstruction of the signal by using a new singular value matrix S to realize secondary noise reduction of the signal.
The third step: extracting an input feature vector of fault diagnosis, wherein the implementation process comprises the following steps:
1. coarse grain treatment
The initial signal (u (i) processed in the two steps, i is 1,2, … L) with the length of L is divided into non-overlapping data with the scale of tau, the coarse grained signal is represented by the average value of each data segment, and the k-th coarse grained sequence is represented by
Figure BDA0002485374940000091
The calculation formula is as follows:
Figure BDA0002485374940000092
2) and calculating the spread entropy values of the coarse grained sequences corresponding to all tau by the following process:
(1) mapping y to y by normal cumulative distribution function method1,y2,…,ym}。
Figure BDA0002485374940000093
Where μ and σ are the mean and standard deviation, respectively, of the signal y.
Mapping y to [1, c ] using a linear algorithm]On a certain integer in the interval, the k-th class of the classified time series is expressed as
Figure BDA0002485374940000094
Figure BDA0002485374940000095
Taking an integer as:
Figure BDA0002485374940000096
(2) calculating an embedding vector, the embedding vector being represented as
Figure BDA0002485374940000097
Figure BDA0002485374940000098
Is composed of a delay time d and an embedding dimension m:
Figure BDA0002485374940000099
(3) c is the classification category and m is the embedding dimension, which can form cmA mode, for each cmPotential dispersion mode of
Figure BDA00024853749400000910
Is assigned to
Figure BDA00024853749400000911
Walking mode of
Figure BDA00024853749400000912
Number of
Figure BDA00024853749400000913
Comprises the following steps:
Figure BDA00024853749400000914
3) the spread entropy value is defined by the entropy of the information as,
Figure BDA0002485374940000101
DE is the entropy of dispersion, which is given in units of nit/nat, ln is the base 10 logarithm.
4) The multi-scale dispersion entropy at scale factor tau is calculated as,
Figure BDA0002485374940000102
MDE is multi-scale dispersion entropy, signals of the vibration state signals of the rolling bearing after combined modal decomposition filtering and singular value decomposition filtering are x, m, c, d and tau which are defined above, and the signals are characterized by the multi-scale dispersion entropy; factors that determine the effect on the multiscale spread entropy are the embedding dimension m, the delay time d and the class c, where m is 2, c is 6, d is 1, and the maximum time interval τ is takenmaxCalculating a multi-scale spreading entropy value of the secondary reconstruction signal as an input feature vector of fault diagnosis, wherein the multi-scale spreading entropy value is 20;
the fourth step: the process of establishing the mixed gauss continuous hidden markov rolling bearing fault diagnosis model is shown in fig. 4:
and (3) processing the bearing fault historical data with the labels through the first step, the second step and the third step, calculating to obtain the multi-scale dispersion entropy of the reconstructed signal as a training sample, and training a Gaussian mixture continuous Markov fault diagnosis model. The vibration signal of the processed rolling bearing is O ═ O1,o2,…,oTThe fault state sequence of the running of the rolling bearing equipment is I ═ I1,i2,…,iN}. Let the parameter of the Gaussian mixture continuous Markov model be lambdaiThe device runs L states corresponding to L models
Figure BDA0002485374940000103
Observable information of computing deviceThe probability of a possible fault condition I under O is:
Figure BDA0002485374940000104
the calculation process is as follows: in the algorithm, the model parameters λ ═ (pi, a, B), where a ═ aijIs the state transition probability distribution, pi ═ piiIs the initial state probability distribution. B is that the observed probability distribution is usually fitted with a gaussian mixture distribution, i.e.:
Figure BDA0002485374940000105
in the formula: m is the number of Gauss elements, cjmThe mixing coefficient is the mth Gaussian element of the jth state; mu.sjmAnd σjmMean vector and covariance matrix of m gaussians in j states.
Forward probability of failure αt(i) The calculation formula of (a) is as follows:
αt(i)＝P(o1,o2…ot,iT＝qi|λ),
otfor bearing running observation sequences, itFor a sequence of bearing operating states, qiThe state sequence value. The recurrence formula is:
Figure BDA0002485374940000111
the backward probability of fault occurrence and its recurrence formula are:
βt(i)＝P(ot+1,ot+2…oT,it＝qi|λ),
Figure BDA0002485374940000112
calculating the probability P (O | λ) in conjunction with the definition of the forward and backward algorithmi) Namely, the probability of the rolling bearing fault is as follows:
Figure BDA0002485374940000113
different fault history data can train a corresponding fault diagnosis model.
The fifth step: and processing the collected vibration signal data of the operating bearing through the first step, the second step and the third step, calculating to obtain the multi-scale dispersion entropy of the reconstructed signal, inputting the multi-scale dispersion entropy into the model established in the fourth step, and determining the operating state of the vibration signal by adopting a maximum likelihood probability comparison method. As shown in fig. 4:
maximum likelihood probability comparison: according to the number i of the training fault samples, i fault diagnosis models, namely lambda, can be trainediCalculating the forward and backward probability of the fault
Figure BDA0002485374940000114
The numerical value of the probability indicates the likelihood of failure. Formula for substituting multi-scale dispersion entropy of new sample
Figure BDA0002485374940000121
Calculating the maximum likelihood occurrence probability at λiThe model with the highest probability of occurrence is determined as the fault in the (1,2, …, n) models.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (1)

1. A rolling bearing fault diagnosis system characterized in that:
the first step is as follows: the combined modal decomposition primary filtering of the fault signals of the rolling bearing comprises the following steps:
1) selecting vibration signals of a vibration sensor of the bearing and carrying out signal grouping, wherein the signal grouping is to divide time continuous signals in a time period and calculate the arrangement entropy of each group of original vibration signals;
a. if the permutation entropy is less than 0.618, selecting intrinsic empirical mode decomposition to carry out original signal decomposition, decomposing to obtain intrinsic modal components and residual amount, and entering b;
if the permutation entropy is larger than 0.618, carrying out complete set empirical mode decomposition on the self-adaptive noise to obtain a natural modal component and a residual amount, and entering b;
b. the number of the extreme value points of the residual quantity is judged,
if the number of the extreme points is more than or equal to 3, repeating the operation a to obtain the next inherent modal component and residual quantity;
if the number of the extreme values is less than 3, adding all the inherent modal components obtained by decomposition to obtain the inherent modal component after the initial filtering of the original signal;
2) selecting inherent modal components of an original signal after primary filtering, calculating a correlation coefficient and a kurtosis value, and selecting the inherent modal components of which the correlation coefficient is greater than 0.1 and the kurtosis is greater than 3; sequentially superposing the intrinsic modal components according to the kurtosis values from large to small, and calculating the kurtosis value of a superposed signal once every superposition; finally, selecting the superposed signal with the maximum kurtosis value as a final reconstruction signal;
the second step is that: the second filtering of the signal by singular value decomposition is carried out as follows:
constructing the vibration signal with the length of N and after filtering output in the first step into an m-row N-column Hankel matrix, establishing rules that the row number m is N/2(N is an even number) or (N +1)/2(N is an odd number) and the column number N is N +1-m, m is more than or equal to 2, N is more than 2 and m is less than N, performing singular value decomposition on the Hankel matrix, determining the effective order of singular values by using a singular value difference spectrum maximum method to obtain a new singular value matrix S' reflecting the signal characteristics, and performing singular value decomposition inverse operation according to the matrix to obtain a secondary reconstruction signal;
the third step: and extracting an input feature vector of fault diagnosis, wherein the process is as follows:
after the vibration state signal of the rolling bearing is subjected to combined modal decomposition filtering and singular value decomposition filtering, performing characteristic representation by using a multi-scale diffusion entropy; factors that determine the effect on the multiscale spread entropy are the embedding dimension m, the delay time d, and the class c, where m is 2, c is 6, d is 1, and the maximum is takenTime interval taumaxCalculating a multi-scale spreading entropy value of the secondary reconstruction signal as an input feature vector of fault diagnosis, wherein the multi-scale spreading entropy value is 20;
the fourth step: establishing a Gaussian mixture continuous hidden Markov rolling bearing fault diagnosis model:
processing labeled bearing fault historical data through a first step, a second step and a third step, calculating to obtain multi-scale dispersion entropy of a reconstructed signal as a training sample, and calculating model parameters lambda (pi, A and B) of a Gaussian mixture continuous hidden Markov model, wherein A is { a ═ BijIs the state transition probability distribution, pi ═ piiB, fitting the probability distribution of the observed values by using Gaussian mixture distribution to establish a rolling bearing fault diagnosis model;
the fifth step: and processing the collected vibration signal data of the operating bearing through the first step, the second step and the third step, calculating to obtain the multi-scale dispersion entropy of the reconstructed signal, inputting the multi-scale dispersion entropy into the model established in the fourth step, and determining the operating state of the vibration signal by adopting a maximum likelihood probability comparison method.
CN202010389878.4A 2020-05-11 2020-05-11 Rolling bearing fault diagnosis system Expired - Fee Related CN111444988B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010389878.4A CN111444988B (en) 2020-05-11 2020-05-11 Rolling bearing fault diagnosis system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010389878.4A CN111444988B (en) 2020-05-11 2020-05-11 Rolling bearing fault diagnosis system

Publications (2)

Publication Number Publication Date
CN111444988A true CN111444988A (en) 2020-07-24
CN111444988B CN111444988B (en) 2021-09-17

Family

ID=71657608

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010389878.4A Expired - Fee Related CN111444988B (en) 2020-05-11 2020-05-11 Rolling bearing fault diagnosis system

Country Status (1)

Country Link
CN (1) CN111444988B (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111855192A (en) * 2020-07-31 2020-10-30 北京航空航天大学 Singular value decomposition method for denoising encoder signal
CN112731098A (en) * 2020-12-03 2021-04-30 西安电子科技大学 Radio frequency low-noise discharge circuit fault diagnosis method, system, medium and application
CN113470694A (en) * 2021-04-25 2021-10-01 重庆市科源能源技术发展有限公司 Remote listening monitoring method, device and system for hydraulic turbine set
CN113591248A (en) * 2021-08-09 2021-11-02 兰州理工大学 Bearing fault diagnosis method in mine hoist transmission part
CN113673468A (en) * 2021-08-30 2021-11-19 武汉理工大学 Conveyor fault diagnosis method and system, electronic equipment and storage medium
WO2022037068A1 (en) * 2021-03-22 2022-02-24 南京航空航天大学 Method for diagnosis of fault in machine tool bearing
CN114088400A (en) * 2021-11-01 2022-02-25 中国人民解放军92728部队 Rolling bearing fault diagnosis method based on envelope permutation entropy
CN114167838A (en) * 2021-12-03 2022-03-11 西安电子科技大学 Multi-scale health assessment and fault prediction method for servo system

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102854015A (en) * 2012-10-15 2013-01-02 哈尔滨理工大学 Diagnosis method for fault position and performance degradation degree of rolling bearing
CN107817106A (en) * 2017-10-11 2018-03-20 温州大学 Fault Diagnosis of Roller Bearings based on Bayes's residual transform singular value decomposition Gaussian Mixture HMM framework
CN109583386A (en) * 2018-11-30 2019-04-05 中南大学 A kind of intelligent rotating mechanical breakdown depth network characterization discrimination method
CN109612732A (en) * 2019-01-24 2019-04-12 北华大学 A kind of Fault Diagnosis of Roller Bearings based on deconvolution and envelope spectrum

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102854015A (en) * 2012-10-15 2013-01-02 哈尔滨理工大学 Diagnosis method for fault position and performance degradation degree of rolling bearing
CN107817106A (en) * 2017-10-11 2018-03-20 温州大学 Fault Diagnosis of Roller Bearings based on Bayes's residual transform singular value decomposition Gaussian Mixture HMM framework
CN109583386A (en) * 2018-11-30 2019-04-05 中南大学 A kind of intelligent rotating mechanical breakdown depth network characterization discrimination method
CN109612732A (en) * 2019-01-24 2019-04-12 北华大学 A kind of Fault Diagnosis of Roller Bearings based on deconvolution and envelope spectrum

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
HUIMING JIANG等: "Study on Hankel matrix-based SVD and its application in rolling element bearing fault diagnosis", 《MECHANICAL SYSTEMS AND SIGNAL PROCESSING》 *

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111855192A (en) * 2020-07-31 2020-10-30 北京航空航天大学 Singular value decomposition method for denoising encoder signal
CN111855192B (en) * 2020-07-31 2021-04-23 北京航空航天大学 Singular value decomposition method for denoising encoder signal
CN112731098A (en) * 2020-12-03 2021-04-30 西安电子科技大学 Radio frequency low-noise discharge circuit fault diagnosis method, system, medium and application
CN112731098B (en) * 2020-12-03 2022-04-29 西安电子科技大学 Radio frequency low-noise discharge circuit fault diagnosis method, system, medium and application
WO2022037068A1 (en) * 2021-03-22 2022-02-24 南京航空航天大学 Method for diagnosis of fault in machine tool bearing
CN113470694A (en) * 2021-04-25 2021-10-01 重庆市科源能源技术发展有限公司 Remote listening monitoring method, device and system for hydraulic turbine set
CN113591248A (en) * 2021-08-09 2021-11-02 兰州理工大学 Bearing fault diagnosis method in mine hoist transmission part
CN113673468A (en) * 2021-08-30 2021-11-19 武汉理工大学 Conveyor fault diagnosis method and system, electronic equipment and storage medium
CN114088400A (en) * 2021-11-01 2022-02-25 中国人民解放军92728部队 Rolling bearing fault diagnosis method based on envelope permutation entropy
CN114088400B (en) * 2021-11-01 2024-04-09 中国人民解放军92728部队 Rolling bearing fault diagnosis method based on envelope permutation entropy
CN114167838A (en) * 2021-12-03 2022-03-11 西安电子科技大学 Multi-scale health assessment and fault prediction method for servo system

Also Published As

Publication number Publication date
CN111444988B (en) 2021-09-17

Similar Documents

Publication Publication Date Title
CN111444988B (en) Rolling bearing fault diagnosis system
CN108231201B (en) Construction method, system and application method of disease data analysis processing model
CN109916628B (en) Rolling bearing fault diagnosis method based on improved multi-scale amplitude perception permutation entropy
CN109000930B (en) Turbine engine performance degradation evaluation method based on stacking denoising autoencoder
CN109389171B (en) Medical image classification method based on multi-granularity convolution noise reduction automatic encoder technology
CN108399248A (en) A kind of time series data prediction technique, device and equipment
CN113869208B (en) Rolling bearing fault diagnosis method based on SA-ACWGAN-GP
CN105116397B (en) Radar high resolution range profile target identification method based on MMFA models
CN109496334A (en) For assessing the device and method of voice quality
CN110705525A (en) Method and device for diagnosing rolling bearing fault
CN109443752B (en) Gear vibration signal noise reduction and fault diagnosis method based on VMD
CN105260998A (en) MCMC sampling and threshold low-rank approximation-based image de-noising method
CN110795843A (en) Method and device for identifying faults of rolling bearing
CN113435322A (en) Main shaft bearing fault detection method, system, equipment and readable storage medium
CN102901630A (en) Adaptive redundant lifting wavelet noise reduction analysis-based bearing failure recognition method
CN115905976B (en) Attention mechanism-based fault diagnosis method, system and equipment for Highway Bi-LSTM bearing
CN115587666A (en) Load prediction method and system based on seasonal trend decomposition and hybrid neural network
Zhu et al. Deep fourier kernel for self-attentive point processes
CN116337449A (en) Sparse self-coding fault diagnosis method and system based on information fusion
Lee et al. An explainable deep learning-based prognostic model for rotating machinery
CN114264478A (en) Diesel engine crankshaft bearing wear degree prediction method and system
CN116541771A (en) Unbalanced sample bearing fault diagnosis method based on multi-scale feature fusion
CN114383845A (en) Bearing composite fault diagnosis method based on embedded zero sample learning model
CN113435321A (en) Method, system and equipment for evaluating state of main shaft bearing and readable storage medium
CN113449618A (en) Method for carrying out deep learning rolling bearing fault diagnosis based on feature fusion and mixed enhancement

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20210917

CF01 Termination of patent right due to non-payment of annual fee