CN114608826A - Training method, diagnosis method and diagnosis device of bearing fault diagnosis model - Google Patents

Training method, diagnosis method and diagnosis device of bearing fault diagnosis model Download PDF

Info

Publication number
CN114608826A
CN114608826A CN202210259973.1A CN202210259973A CN114608826A CN 114608826 A CN114608826 A CN 114608826A CN 202210259973 A CN202210259973 A CN 202210259973A CN 114608826 A CN114608826 A CN 114608826A
Authority
CN
China
Prior art keywords
fault
bearing
training
signal
fault diagnosis
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.)
Pending
Application number
CN202210259973.1A
Other languages
Chinese (zh)
Inventor
陈剑
季磊
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.)
Hefei University of Technology
Original Assignee
Hefei University of Technology
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 Hefei University of Technology filed Critical Hefei University of Technology
Priority to CN202210259973.1A priority Critical patent/CN114608826A/en
Publication of CN114608826A publication Critical patent/CN114608826A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • 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
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
    • 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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • 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
    • G06F2218/04—Denoising
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Acoustics & Sound (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

The invention provides a training method, a diagnosis method and a diagnosis device of a bearing fault diagnosis model, wherein the training method comprises the following steps: collecting fault vibration signals of different types of faults of a bearing to obtain a sample set; performing empirical Fourier decomposition on the fault vibration signals in the sample set to obtain a plurality of modal signal components; screening a plurality of modal signal components through kurtosis and correlation coefficients respectively, and reconstructing screening results to obtain a reconstructed signal sequence; processing the reconstructed signal sequence by utilizing an improved hierarchical slope entropy algorithm to obtain fault characteristic information and form a training characteristic set; and training the bearing fault diagnosis model by using the training feature set to obtain the trained bearing fault diagnosis model. The method can effectively identify the bearing faults of different types and damage degrees, and the average identification rate reaches 99.74%.

Description

Training method, diagnosis method and diagnosis device of bearing fault diagnosis model
Technical Field
The invention belongs to the technical field of bearing fault diagnosis, and particularly relates to a training method, a diagnosis method and a diagnosis device of a bearing fault diagnosis model.
Background
Mechanical equipment is developing towards large-scale and precise direction, the intellectualization is increasing day by day, and higher requirements are put forward for the reliability of the equipment in operation. As a core slewing bearing component in a rotary machine, a slight defect in the surface thereof may cause an operational failure of the entire installation system, resulting in huge casualties and property loss. The traditional rolling bearing detection method needs workers to disassemble the bearing from mechanical equipment regularly for safety inspection, consumes a large amount of manpower and material resources, and affects the industrial production process. The intelligent maintenance scheme utilizes advanced sensors and detection technology to obtain the running state information of the equipment, when a fault occurs, the residual service life of the equipment can be predicted by establishing a physical and statistical model, and the equipment is maintained before the equipment is completely failed. Therefore, bearing fault diagnosis and state detection are carried out, and the bearing fault diagnosis and state detection method has important significance for improving the operation safety of equipment.
The bearing is used as a key precise element of a rotating machine, and various complex working conditions can cause the bearing to be damaged by fatigue spalling, cracks, abrasion, indentation and the like, so that the vibration of the bearing is aggravated, and potential safety hazards are generated. In order to reduce the occurrence of accidents, it is necessary to perform condition monitoring and fault diagnosis on the bearings.
Disclosure of Invention
In order to solve the technical problems, the invention provides a training method, a diagnosis method and a diagnosis device for a bearing fault diagnosis model, which are used for realizing state monitoring and fault diagnosis of a bearing, so that the bearing is prevented from being damaged by fatigue spalling, cracks, abrasion, indentation and the like to aggravate bearing vibration, and potential safety hazards are avoided.
The invention provides a training method of a bearing fault diagnosis model, which comprises the following steps:
collecting fault vibration signals of different types of faults of a bearing to obtain a sample set;
performing empirical Fourier decomposition on the fault vibration signals in the sample set to obtain a plurality of modal signal components;
screening a plurality of modal signal components through kurtosis and correlation coefficients respectively, and reconstructing screening results to obtain a reconstructed signal sequence;
processing the reconstructed signal sequence by utilizing an improved hierarchical slope entropy algorithm to obtain fault characteristic information and form a training characteristic set;
and training the bearing fault diagnosis model by using the training feature set to obtain the trained bearing fault diagnosis model.
In one embodiment of the invention, the different types of faults include at least normal, inner ring single point fault, outer ring single point fault, roller single point fault, outer ring roller composite fault, inner ring multipoint fault, outer ring multipoint fault and ball multipoint fault.
In one embodiment of the present invention, the step of performing an empirical fourier decomposition on the sample set to obtain the plurality of modal signal components comprises:
normalizing the Fourier spectrum of the fault vibration signal in the sample set to a [0, pi ] interval, and predefining the frequency band number N required to be divided;
setting initial values and maximum values in a Fourier spectrum as M control points, and acquiring a final control point number T according to the frequency band number N and the control point number M;
dividing intervals according to the final control point number T, and the position of each control point is alphanWherein n is more than or equal to 1 and less than or equal to T and alpha1＝0，αT+1＝π；
According to
Figure BDA0003550397690000021
Determining spectral segmentation boundaries omega for Fourier decompositionnDefining the first T boundaries as min [ alpha ]T-1，αT]Wherein ΛnIs a set of the first T boundaries;
determining a continuous interval omega-omegai，ωi+1]And calculating an analytic Fourier eigenband function of each interval, wherein i is 1,2, …, T;
and carrying out inverse Fourier transform on the real part of the analyzed Fourier inherent frequency band function of each interval to obtain a plurality of modal signal components.
In an embodiment of the present invention, obtaining the final control point number T according to the frequency segment number N and the control point number M includes:
when M is larger than or equal to N, only taking the first N segments of the M control points which are arranged in a descending order, namely the number T of the final control points is N;
and when M is less than N, the frequency bands capable of being divided in the vibration signals are less than the predefined frequency band number N, and the final control point number T is M.
In one embodiment of the invention, screening a plurality of the modal signal components by kurtosis and correlation coefficients to obtain a reconstructed signal comprises:
calculating Kurt value of each modal signal componentnAnd retaining Kurt valuenAll components greater than 3 to form a set U;
calculating a correlation coefficient rho of each modal signal component and the fault vibration signal ynAnd by a correlation coefficient rhonThe standard deviation of the mode signal is the lowest threshold value, and each mode signal component is screened to be used as the main rotation component of the fault vibration signal to form a set V;
and taking the union of the set U and the set V and adding signals therein to obtain a reconstructed signal sequence.
In one embodiment of the present invention, the kurtosis value KurtnBy the formula:
Figure BDA0003550397690000031
calculated to obtain wherein efdn,kK points for the nth modal signal component; n-1, 2, …, T; k is 1,2, …, L; u and σ are efdnMean and variance of; e [ n ]]Representing the mathematical expectation of n.
In one embodiment of the invention, the correlation coefficient ρ isnBy the formula:
Figure BDA0003550397690000032
and calculating to obtain the result, wherein,
Figure BDA0003550397690000033
represents the mean value of the n-th modal component, zkThe kth sample point representing the fault vibration signal z,
Figure BDA0003550397690000034
representing the mean of all sample points of the fault vibration signal z.
In one embodiment of the present invention, processing the reconstructed signal sequence by using an improved hierarchical slope entropy algorithm to obtain fault feature information and form a training feature set includes:
given a reconstructed signal sequence x ═ xiI 1,2, …, L, and calculating the average operator P0And difference operator P1：
Figure BDA0003550397690000035
Wherein x isiA reconstructed signal representing an i-th section;
defining operators at different layer numbers s
Figure BDA0003550397690000036
(f ═ 0 or 1), wherein,
Figure BDA0003550397690000037
given a particular sequence of vectors r1,r2,…,rs]And determining the f value of each layer according to the node e of the s-th layer, wherein the relationship between the node e and the vector sequence is as follows:
Figure BDA0003550397690000038
wherein, rdD ═ 1,2, …, s } - [0,1] denotes an average operator or a difference operator at the d-th layer;
the hierarchical components of the reconstructed signal sequence are obtained, expressed as:
Figure BDA0003550397690000041
calculating an improved level slope entropy for each level component, expressed as: MHSE (X, m, s, e, γ, δ) SE (X)s,eM, γ, δ) where m is the embedding dimension and γ and δ are the threshold values of the slope entropy.
In an embodiment of the present invention, the step of training the bearing fault diagnosis model by using the training feature set to obtain a trained bearing fault diagnosis model includes:
the training feature set is denoted as { u }iI ═ 1,2, …, p }, and the weight between the input layer and the hidden layer is defined as
Figure BDA0003550397690000045
The weight between the hidden and output layers is β, the hidden layer threshold is b, p is 2s*(m-1)：
Figure BDA0003550397690000042
b＝[b1,b2,…,bQ]TR, Q, F represents the number of neurons in the input, hidden, and output layers of the extreme learning machine;
setting an activation function g (x), the output T of the extreme learning machine*Expressed as: t is*H β, wherein
Figure BDA0003550397690000043
Solving the minimum value of the approximate square difference to obtain the weight of the output layer as
Figure BDA0003550397690000044
And solving the optimal solution as beta*＝H+T*In the formula, H+The generalized inverse of the output matrix for the hidden layer.
The invention also provides a bearing fault diagnosis method, which comprises the following steps:
collecting a fault vibration signal of a bearing;
performing empirical Fourier decomposition on the fault vibration signal to obtain a plurality of modal signal components;
screening a plurality of modal signal components through kurtosis and correlation coefficients respectively, and reconstructing screening results to obtain a reconstructed signal sequence;
processing the reconstructed signal sequence by using an improved hierarchical slope entropy algorithm to obtain fault characteristic information;
inputting the fault characteristic information into a bearing fault diagnosis model trained by the bearing fault diagnosis model training method according to any one of claims 1 to 9 to obtain a diagnosis result of the bearing fault.
The invention also provides a bearing fault diagnosis device, which comprises:
the fault signal acquisition module is used for acquiring fault vibration signals of different types of faults of the bearing so as to obtain a sample set;
the modal signal acquisition module is connected with the fault signal acquisition module and is used for performing empirical Fourier decomposition on the fault vibration signals in the sample set to acquire a plurality of modal signal components;
the signal reconstruction module is connected with the modal signal acquisition module and is used for screening a plurality of modal signal components through kurtosis and correlation coefficients respectively and reconstructing a screening result to acquire a reconstruction signal sequence;
the training feature set acquisition module is connected with the signal reconstruction module and used for processing the reconstructed signal sequence by utilizing an improved hierarchical slope entropy algorithm to acquire fault feature information and form a training feature set;
the model construction module is connected with the training feature set acquisition module and used for training the bearing fault diagnosis model by using the training feature set so as to acquire a trained bearing fault diagnosis model;
and the fault diagnosis module is connected with the model construction module and is used for acquiring a diagnosis result.
The invention provides a training method, a diagnosis method and a diagnosis device of a bearing fault diagnosis model, which can be used for preprocessing a rolling bearing fault signal by utilizing empirical Fourier decomposition, filtering a noise section of a Fourier frequency spectrum and highlighting a main vibration component and an impact component of a time sequence.
Compared with the method for coarse graining sequence, the improved hierarchical slope entropy algorithm provided by the invention has the advantages that the extracted time sequence scale information is more comprehensive and abundant; meanwhile, the time series difference is measured by the slope, the trend and the degree of amplitude change can be sensed, and the method of sorting the relative arrangement Entropy (PE) by the amplitude is more effective.
The training method, the diagnosis method and the diagnosis device of the bearing fault diagnosis model provided by the invention can effectively identify bearing faults of different types and damage degrees, and the average identification rate reaches 99.74%.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a flow chart of a bearing fault diagnosis method according to the present invention.
Fig. 2 is a flowchart of a bearing fault diagnosis method according to the present invention.
Fig. 3 is a time series difference versus symbol.
Fig. 4 is a symbol sequence pattern when the embedding dimension m is 3.
Fig. 5 is a schematic diagram of a level decomposition in the bearing fault diagnosis method provided by the present invention.
Fig. 6 is a schematic diagram of a bearing fault diagnosis device according to the present invention.
Fig. 7 is a waveform diagram of the first 5 components obtained by performing modal decomposition on the outer ring multi-point fault according to an embodiment of the present invention.
FIG. 8 is a t-SNE visual dimension reduction result of MHSE features of 9 sets of fault samples according to an embodiment of the present invention.
FIG. 9 is a test set classification confusion matrix according to an embodiment of the present invention.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention.
It should be noted that the drawings provided in the present embodiment are only for illustrating the basic idea of the present invention, and the components related to the present invention are only shown in the drawings rather than drawn according to the number, shape and size of the components in actual implementation, and the type, quantity and proportion of the components in actual implementation may be changed freely, and the layout of the components may be more complicated.
The invention provides a training method, a diagnosis method and a diagnosis device of a bearing fault diagnosis model, which are used for realizing state monitoring and fault diagnosis of a bearing, so that the bearing is prevented from being damaged by fatigue peeling, cracks, abrasion, indentation and the like to aggravate bearing vibration and further avoid potential safety hazards, and concretely, as shown in figures 1 and 2, the diagnosis method comprises the following steps:
s1, collecting fault vibration signals of different types of faults of the bearing to obtain a sample set;
s2, carrying out empirical Fourier decomposition on the fault vibration signals in the sample set to obtain a plurality of modal signal components;
s3, screening the modal signal components through kurtosis and correlation coefficients respectively, and reconstructing screening results to obtain a reconstructed signal sequence;
s4, processing the reconstructed signal sequence by using an improved hierarchical slope entropy algorithm to obtain fault feature information and form a training feature set;
s5, training the bearing fault diagnosis model by using the training feature set to obtain a trained bearing fault diagnosis model;
as shown in fig. 1 and 2, in step S1, different types of fault vibration signals of the bearing are collected to obtain a sample set, and the sample set is used for training and obtaining a diagnostic model. In this embodiment, the different types of faults at least include normal, inner ring single point fault, outer ring single point fault, roller single point fault, outer ring roller composite fault, inner ring multipoint fault, outer ring multipoint fault, and ball multipoint fault.
As shown in fig. 1 and fig. 2, in step S2, an empirical fourier decomposition is performed on the faulty vibration signals in the sample set to obtain a plurality of modal signal components, wherein the empirical fourier decomposition is performed by dividing frequency bands by a modified spectrum division technique, and then a zero-phase filter is constructed for filtering, and each frequency band is subjected to an inverse fourier transform to obtain a vibration signal decomposition component. Specifically, the step of obtaining a plurality of modal signal components includes:
s21, carrying out Fourier spectrum normalization on the fault vibration signals in the sample set to be in a [0, pi ] interval, and predefining the frequency band number N required to be divided;
s22, determining M control points according to initial values and maximum values in the Fourier spectrum, and acquiring final control point numbers T according to the frequency segment number N and the control point numbers M;
s23, dividing the interval according to the final control point number T, and setting the position of each control point as alphanWherein n is more than or equal to 1 and less than or equal to T and alpha1＝0，αT+1＝π；
S24, according to
Figure BDA0003550397690000071
Determining spectral segmentation boundaries omega for Fourier decompositionnDefining the first T boundaries as min [ alpha ]T-1，αT]Wherein Λ isnIs a set of the first T boundaries;
s25, determining a continuous interval omega ═ omegai，ωi+1]And calculating an analytic Fourier eigenband function of each interval, wherein i is 1,2, …, T;
s26, performing inverse fourier transform on the real part of the analytic fourier eigenband function of each interval to obtain a plurality of modal signal components.
As shown in fig. 1 and fig. 2, the relationship between the final control point number T and the number N of frequency segments and the control point number M is: when M is larger than or equal to N, only taking the first N segments of the M control points which are arranged in a descending order, namely the number T of the final control points is N; and when M is less than N, the frequency range which can be divided in the vibration signal is less than the predefined frequency range number N, and the final control point number T is M.
As shown in fig. 1, the interval is divided according to the final control point number T, and the position of each control point is alphanWherein n is more than or equal to 1 and less than or equal to T and alpha1＝0，αT+1Pi; it should be noted that T +1 boundaries are required for T intervals, and the first T boundaries min [ α ] are definedT-1，αT]The set of the first T boundaries is ΛnThe T +1 th boundary is alphaTAnd alphaT+1The mid-point of (c), the spectral segmentation boundary ω of the fourier decomposition can be determinedn，
Figure BDA0003550397690000072
In step S25, the continuous section Ω ═ ω is determinedi，ωi+1]And calculating an analytic fourier eigenband function for each interval, where i is 1,2, …, T. In particular, assume x [ n ]]Is a discrete continuous sequence of length L whose discrete Fourier form can be written as:
Figure BDA0003550397690000073
in the formula,
Figure BDA0003550397690000074
is the sequence X [ n ]]Given that L is an even number, X0]And
Figure BDA0003550397690000075
are all real numbers. It is rewritten as:
Figure BDA0003550397690000081
in the formula, Re { z1[n]Denotes z1[n]Real part of (z)1[n]And z2[n]Are complex conjugates of each other. Analysis of the signal z1[n]The following can be written:
Figure BDA0003550397690000082
calculating an analytic fourier eigenband function for each interval:
Figure BDA0003550397690000083
wherein i is 1,2, …, T.
In step S26, after performing inverse fourier transform on the real part of the analyzed fourier-eigenband function for each interval, a plurality of modal signal components (efd components) corresponding to each interval are acquired.
As shown in fig. 1 and fig. 2, in step S3, the method includes screening a plurality of modal signal components by kurtosis and correlation coefficients, and reconstructing a screening result to obtain a reconstructed signal sequence, specifically including:
s31, calculating Kurt value of each modal signal componentnAnd retaining Kurt valuenAll components greater than 3 to form a set U, wherein the kurtosis value KurtnBy the formula:
Figure BDA0003550397690000084
calculated to obtain wherein efdn,kK points for the nth modal signal component; n-1, 2, …, T; k is 1,2, …, L; u and σ are efdnMean and variance of; e [ n ]]Representing the mathematical expectation of n.
S32, calculating a correlation coefficient rho of each modal signal component and the fault vibration signal ynAnd are related toCoefficient ρnIs the lowest threshold value, and screening each modal signal component as the main rotation component of the fault vibration signal to form a set V, wherein the correlation coefficient rho isnBy the formula:
Figure BDA0003550397690000085
and calculating to obtain the result, wherein,
Figure BDA0003550397690000086
represents the mean value of the n-th modal component, zkThe kth sample point representing the fault vibration signal z,
Figure BDA0003550397690000091
representing the mean of all sample points of the fault vibration signal z.
And S33, taking the union of the set U and the set V and adding the signals therein to obtain a reconstructed signal sequence.
Aiming at the defect that the utilization of the permutation Entropy to the amplitude information lacks completeness, Slope Entropy (SE) based on symbol and amplitude information is provided, SE adopts a similar linear quantization mode, and slopes between sequence elements are divided into different modes by defining a threshold interval, so as to measure the change trend of the amplitude, specifically, the method comprises the following steps:
first, assume a time series y ═ yiI-1, 2, …, L }, and the subsequence can be obtained by reconstructing the phase space of the i-1, 2, … and L }, and then obtaining the subsequence
Figure BDA0003550397690000092
Figure BDA0003550397690000093
Where t is 1,2, …, L-m +1, m being the embedding dimension. A vertical increment threshold gamma and a horizontal increment threshold delta are defined. Gamma is a larger quantity used for measuring the obvious difference between vector sequences so as to distinguish different fluctuation amplitudes; δ is a very small value, so as to classify the case of approximately equal magnitude. Secondly, the time sequence is differentiated, and the sub-sequences are processed by a threshold valueSequences are defined as different symbols and figure 3 shows the time series difference versus symbol.
When y ist+τ-yτ>When γ, the symbol is defined as + 2.
When y ist+τ-yτIs not less than delta and yt+τ-yτ<When γ, the symbol is defined as + 1. When γ is 1, the change in slope at this time corresponds to a range of [0,45 °).
When yt+τ-yτWhen | ≦ γ, the symbol is defined as 0 in the region close to the 0 difference.
When y ist+τ-yτIs not more than-delta and yt+τ-yτ>When- γ, the symbol is defined as-1. When γ is 1, the slope change at this time corresponds to a range of [ -45 °, 0).
When y ist+τ-yτ<When- γ, the symbol is defined as-2.
The time sequence is defined as above to obtain a symbol sequence Sm＝{siI ═ 1,2 …, L-1}, the symbolized subspace can be expressed as:
Figure BDA0003550397690000094
wherein L is 1,2, …, L-m + 2. Subspaces below the dimension ε -m-1 are also counted, the obtainable dimensions d 1,2 …, respectively, and the sequence patterns formed are written as:
Figure BDA0003550397690000095
wherein
Figure BDA0003550397690000096
Are subspaces of different dimensions. Fig. 4 shows the symbol sequence pattern when the embedding dimension m is 3. Counting the total number Z of the arrangement modes appearing in each sequence modedThe number of occurrences of each permutation is hd,iThe corresponding probability p can be obtainedd,i＝hd,i/ZdWhere i is 1,2, …, Zd. Defining shannon entropy for each sequence pattern:
Figure BDA0003550397690000097
the slope entropy can be expressed as:
SE＝{sed,d＝1,2,…,ε}。
in this embodiment, as shown in fig. 1 to fig. 5, in step S4, it should be noted that the Hierarchical Entropy (HE) obtains implicit high and low frequency information by an average and difference method, but the hierarchical process thereof needs to ensure that the data length L is 2n(n is a positive integer). Meanwhile, the hierarchical entropy gradually loses statistical significance along with the increase of the hierarchy, and the reliability of decomposition is reduced. Specifically, the processing of the reconstructed signal sequence by using an improved hierarchical slope entropy algorithm to obtain fault feature information and form a training feature set includes:
s41, providing a reconstructed signal sequence x ═ { x ═ xiI 1,2, …, L, and calculating the average operator P0And difference operator P1：
Figure BDA0003550397690000101
Wherein x isiA reconstructed signal representing an i-th section;
s42, defining operators at different layer numbers S
Figure BDA0003550397690000102
(f ═ 0 or 1), wherein,
Figure BDA0003550397690000103
s43, a specific vector sequence is given1,r2,…,rs]Determining the f value of each layer according to the node e of the s-th layer, and the relation between the node e and the vector sequenceComprises the following steps:
Figure BDA0003550397690000104
wherein, rdD ═ 1,2, …, s } - [0,1] denotes an average operator or a difference operator at the d-th layer; node e may be compared to fig. 3, e starting from 0, when s is 1, there are two nodes 0, 1; when s is 2, there are four nodes 0,1,2, 3.
It should be noted that, in order to obtain the time-series hierarchical information, the operator defined above needs to be repeatedly used
Figure BDA0003550397690000105
Specific vector sequence formula
Figure BDA0003550397690000106
Determining, for example, the fourth hierarchical component X of the second layer2,3Wherein 2 represents s ═ 2, i.e., the second layer; 3 denotes e-3, i.e. the 4 th component starting from 0 to 3, i.e. the 4 th component in the second layer, and is represented by the formula:
Figure BDA0003550397690000107
the following can be obtained:
Figure BDA0003550397690000108
x is actually the original sequence y, rdThat is, f can only take 0 or 1. So find X2,3The process comprises the following steps: 21*r1+20*r2＝3，(r1,r2E {0,1}), when r1＝1，r2This equation holds true for 1.
Determining a vector sequence corresponding to each node according to the node e of each layer, wherein each node of different layers has a corresponding vector sequence [ r1,r2,…,rs]E.g. X2,3Vector sequence of [11 ]]I.e. r1＝1,r2＝1；X3,3Vector sequence of [110 ]]I.e. r1＝1,r2＝1,r3I.e. e for each layer determines all f values from the first layer to the current layer.
S44, obtainingThe hierarchical components of the reconstructed signal sequence are represented as:
Figure BDA0003550397690000111
s45, calculating the improved level slope entropy of each level component, and expressing as: MHSE (X, m, s, e, γ, δ) SE (X)s,eM, γ, δ), where m is the embedding dimension m, and γ and δ are thresholds of slope entropy, where some parameters of the MHSE need to be determined artificially, i.e., embedding dimension m, number of decomposition layers s, and thresholds γ and δ, respectively, where the embedding dimension m is taken to be m and the embedding dimension δ is taken to be m in order to ensure that the MHSE contains at least two symbol modes>2, m is 3, and since the number of decomposition layers is too large, the calculation is time-consuming, and s is usually set to 3. The thresholds γ and δ determine the classification interval of the sequence slope change, where γ is 1 and δ is 0.001, and the improved hierarchical slope entropy is the training feature set.
As shown in fig. 1, in step S5, a step of training the bearing fault diagnosis model by using the training feature set to obtain a trained bearing fault diagnosis model, for example, a step of training by using an extreme learning machine to construct a diagnosis model, that is, inputting the training feature set into the extreme learning machine to obtain the diagnosis model. The step of obtaining the trained bearing fault diagnosis model comprises the following steps:
the training feature set is denoted as { u }iI ═ 1,2, …, p }, and the weight between the input layer and the hidden layer is defined as
Figure BDA0003550397690000118
The weight between the hidden and output layers is β, the hidden layer threshold is b, p is 2s*(m-1)：
Figure BDA0003550397690000112
b＝[b1,b2,…,bQ]TIn the formula, R, Q, F represents the numbers of neurons in the input layer, hidden layer and output layer of the extreme learning machine, respectively, and the value of p is related to the obtained level s and the embedding dimension m of the level entropy, that is, p is 2s*(m-1)；
Setting upAn activation function g (x) for outputting T of extreme learning machine*Expressed as: t is*H β, wherein
Figure BDA0003550397690000113
Note that the input layer is multiplied by a weight
Figure BDA0003550397690000114
Adding b, fitting by an activation function g (x) to obtain a hidden layer H, wherein,
Figure BDA0003550397690000115
in [ -1,1 [)]B is [0,1]]And randomly taking values. The activation function g (x) may be a Sigmoid function, a Tanh function, a sin function, etc., such as a Sigmoid function, which is expressed by the following expression:
Figure BDA0003550397690000116
any non-linear function can be approximated using the activation function, so that the neural network can be applied to a wide variety of non-linear models.
Solving the minimum value of the approximate square difference to obtain the weight of the output layer as
Figure BDA0003550397690000117
And solving the optimal solution as beta*＝H+T*In the formula, H+The generalized inverse of the output matrix for the hidden layer.
As shown in fig. 1, in this embodiment, the present invention provides a bearing fault diagnosis method, which diagnoses a bearing fault by using a diagnosis model constructed in the above embodiment, and includes:
s61, collecting fault vibration signals of the bearing;
s62, carrying out empirical Fourier decomposition on the fault vibration signal to obtain a plurality of modal signal components;
s63, screening the modal signal components through kurtosis and correlation coefficients respectively, and reconstructing screening results to obtain a reconstructed signal sequence;
s64, processing the reconstructed signal sequence by using an improved hierarchical slope entropy algorithm to obtain fault characteristic information;
and S65, inputting the fault characteristic information into the bearing fault diagnosis model trained by the bearing fault diagnosis model training method to obtain the diagnosis result of the bearing fault.
As shown in fig. 2, in some embodiments, the fault vibration signals collected and the output diagnosis results in the above embodiments may also be constructed as a test set or a test set is additionally constructed, and the test set is used as an input of a diagnosis model to continuously optimize the diagnosis model.
As shown in fig. 6, the present invention further provides a bearing fault diagnosis apparatus, which applies the bearing fault diagnosis method described in the above embodiment, specifically, the diagnosis model includes a fault signal acquisition module 10, a modal signal acquisition module 20, a signal reconstruction module 30, a training feature set acquisition module 40, a model construction module 50, and a fault diagnosis module 60. The fault signal acquisition module 10 is used for acquiring fault vibration signals of different types of faults of the bearing to obtain a sample set; the modal signal acquisition module 20 is connected to the fault signal acquisition module 10, and is configured to perform empirical fourier decomposition on the fault vibration signals in the sample set to acquire a plurality of modal signal components; the signal reconstruction module 30 is connected to the modal signal acquisition module 20, and configured to respectively screen a plurality of modal signal components by kurtosis and correlation coefficients, and reconstruct a screening result to acquire a reconstructed signal sequence; the training feature set acquisition module 40 is connected to the signal reconstruction module 30, and is configured to process the reconstructed signal sequence by using an improved hierarchical slope entropy algorithm to acquire fault feature information and form a training feature set; the model construction module 50 is connected to the training feature set acquisition module 40, and is configured to train the bearing fault diagnosis model by using the training feature set to acquire a trained bearing fault diagnosis model; the fault diagnosis module 60 is connected to the model construction module 50 for obtaining a diagnosis result.
In the present example, as shown in fig. 1, the test bearings used were aero-engine bearings of the type NU1010EM (inner ring removable) and N10-10EM (outer ring removable), and bearing component damage was machined by a laser marking machine and a wire-cut electrical discharge machine. The bearing damage types comprise health faults, single-point faults and multi-point faults of an inner ring, an outer ring and a roller, and composite faults of the outer ring/the roller and the inner ring/the roller, wherein 9 states are respectively marked by numbers 1-9 as different fault categories, and the table 1 shows the corresponding relation of the faults.
TABLE 1 bearing fault type, quantity and position mapping table
Figure BDA0003550397690000131
Lab software is used for collecting experimental data, the sampling frequency is 20480Hz, the bearing working condition is 3kn of axial load, the rotating speed is 3000rpm, all fault signals are subjected to data segmentation according to 1000 sample points of each section, 100 samples are taken for each fault, 20 samples are randomly adopted as a sample set, the rest 80 samples are adopted as a test set, the total number of samples in the sample set is 180, the total number of samples in the test set is 720, and the test set is used for testing a trained diagnostic model.
Firstly, empirical fourier decomposition is carried out on a fault sample signal, and in order to separate out the spectrum section of the noise as much as possible, the number of the predefined frequency sections N is 20. Fig. 7 is the waveform of the first 5 components obtained by EFD decomposition of the outer ring multi-point fault, and the decomposed and reconstructed signal contains the main frequency and the impact frequency of the fault vibration signal. After the rolling bearing fault signal is decomposed and reconstructed, feature extraction is carried out through an MHSE algorithm, and each sample can obtain a 16-dimensional feature vector. FIG. 8 is a t-SNE (t-distributed stored Neighbor Embedding) visualization dimension reduction result of MHSE characteristics of 9 groups of fault samples, wherein numbers 1-9 correspond to different fault type labels respectively. As can be seen from FIG. 8, the class characteristics of the 9 faults are compact, and the distinction degree between the other faults is obvious except that part of the characteristics of the tag 5 is closer to the tag 7. The MHSE algorithm can effectively represent different fault types and damage degrees in a whole view.
Meanwhile, in this embodiment, in order to verify the feature extraction performance of the improved hierarchical slope Entropy algorithm, the improved hierarchical arrangement Entropy (MHPE), the fine Composite multi-scale dispersion Entropy (rounded Composite Multiscale D-dispersion Entropy, RCMDE), the fine Composite multi-scale sample Entropy (rounded Composite Multiscale samlpe Entropy, RCMSE), the fine Composite multi-scale Fuzzy Entropy (rounded Composite Multiscale Fuzzy Entropy, mpe), and the specific parameters of the algorithm cwe are set in table 2.
TABLE 2 concrete parameters of the five algorithms
Figure BDA0003550397690000141
After extracting the reconstructed signal features, the overall clustering characteristics of the different methods were jointly evaluated using the contour coefficients and the Davison Baud Index (DBI). The contour coefficient measures the distance between data by defining the cohesion and dispersion, the value range is distributed in [0,1], the closer the numerical value is to 1, the smaller the dispersion in the class is, the larger the dispersion between the classes is. The DBI is the mean value of the maximum similarity between classes, and generally, the smaller the numerical value is, the larger the difference between classes is, the better the classification effect is, and the evaluation result of the species index and the calculation time of a single sample are shown in table 3.
TABLE 3 evaluation results of two indexes and calculation time of individual sample
Figure BDA0003550397690000142
As can be seen from Table 3, the MHSE in the 6 algorithms has the optimum two indexes, and presents better linear clustering characteristics. The MHSE is second to the RCMDE in the time consumption of sample calculation of the same sampling point, and although the RCMDE has the fastest operation rate, the linear clustering characteristic of the RCMDE is much lower than that of the MHSE.
In addition, in the present embodiment, in order to verify the effect of the rolling bearing fault diagnosis method herein, fault characteristics of 9 health states are input to the extreme learning machine for classification experiments. The activation function of the extreme learning machine adopts a Sigmoid function, and the number of neurons is respectively set to be 16 for an input layer R, 30 for a hidden layer Q and 1 for an output layer F. FIG. 9 is a confusion matrix of one experiment, the abscissa 1-9 represents the predicted 9-bearing state, the ordinate 1-9 represents the actual 9-bearing fault, and the diagonal represents the prediction accuracy of each fault. As can be seen from fig. 9, only 1 of the 720 failure test samples is mistakenly classified into other types, and from the type, the inner ring multi-point failure of the label 7 is diagnosed as the inner ring single-point failure. The classification accuracy of the experiment reaches 99.86%, and the effectiveness of the method is verified.
To further highlight the performance of the methods herein, the MHSE was compared to the 5 algorithms of the feature extraction experiments. Table 4 lists the average accuracy and corresponding standard deviation for 10 experiments per method.
TABLE 4 mean accuracy and corresponding standard deviation for 10 experiments per method
Figure BDA0003550397690000151
As can be seen from Table 4, the average accuracy of MHSE reaches 99.74%, the standard deviation is 0.103, and compared with MHPE, RCMDE, RCMSE, RCMFE and CWMPE, the standard deviation of MHSE is minimum, and the average accuracy is respectively improved by 3.35%, 3.89%, 6.7%, 3.59% and 3.24%. From experimental results, the method can effectively distinguish single type and composite type rolling bearing faults, has good distinguishing degree on the faults with different damage degrees, and has better recognition accuracy and stability than other 5 algorithms.
In summary, the invention provides a training method, a diagnosis method and a diagnosis device for a bearing fault diagnosis model, which can filter noise segments of Fourier frequency spectrum and highlight main vibration components and impact components of a time sequence by utilizing empirical Fourier decomposition to preprocess a rolling bearing fault signal.
Compared with the method for coarse graining sequence, the improved hierarchical slope entropy algorithm provided by the invention has the advantages that the extracted time sequence scale information is more comprehensive and abundant; meanwhile, the time series difference is measured by the slope, the trend and the degree of amplitude change can be sensed, and the method of sorting relative arrangement Entropy (PE) by amplitude is more effective.
The bearing fault diagnosis method provided by the invention can effectively identify the bearing faults of different types and damage degrees, and the average identification rate reaches 99.74%.
The above description is only a preferred embodiment of the present application and a description of the applied technical principle, and it should be understood by those skilled in the art that the scope of the present application is not limited to the technical solution of the specific combination of the above technical features, and also covers other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept, for example, the technical solutions formed by mutually replacing the above technical features (but not limited to) having similar functions disclosed in the present application.
Other technical features than those described in the specification are known to those skilled in the art, and are not described herein in detail in order to highlight the innovative features of the present invention.

Claims (11)

1. A training method of a bearing fault diagnosis model is characterized by comprising the following steps:
collecting fault vibration signals of different types of faults of a bearing to obtain a sample set;
performing empirical Fourier decomposition on the fault vibration signals in the sample set to obtain a plurality of modal signal components;
screening a plurality of modal signal components through kurtosis and correlation coefficients respectively, and reconstructing screening results to obtain a reconstructed signal sequence;
processing the reconstructed signal sequence by utilizing an improved hierarchical slope entropy algorithm to obtain fault characteristic information and form a training characteristic set;
and training the bearing fault diagnosis model by using the training feature set to obtain the trained bearing fault diagnosis model.
2. A training method of a bearing fault diagnosis model according to claim 1, characterized in that the different types of faults include at least normal, inner ring single point fault, outer ring single point fault, roller single point fault, outer ring roller compound fault, inner ring multipoint fault, outer ring multipoint fault and ball multipoint fault.
3. A method of training a bearing fault diagnosis model according to claim 1, wherein the step of performing an empirical fourier decomposition on a sample set to obtain a plurality of modal signal components comprises:
normalizing the Fourier spectrum of the fault vibration signal in the sample set to a [0, pi ] interval, and predefining the frequency band number N to be divided;
setting initial values and maximum values in a Fourier spectrum as M control points, and acquiring final control point numbers T according to the frequency segment number N and the control point numbers M;
dividing intervals according to the final control point number T, and the position of each control point is alphanWherein n is more than or equal to 1 and less than or equal to T and alpha1＝0，αT+1＝π；
According to
Figure FDA0003550397680000011
Determining spectral segmentation boundaries omega for Fourier decompositionnDefining the first T boundaries as min [ alpha ]T-1，αT]Wherein ΛnIs a set of the first T boundaries;
determining a continuum omega-omegai，ωi+1]And calculating an analytic Fourier eigenband function of each interval, wherein i is 1,2, …, T;
and carrying out inverse Fourier transform on the real part of the analyzed Fourier inherent frequency band function of each interval to obtain a plurality of modal signal components.
4. The training method of the bearing fault diagnosis model according to claim 3, wherein obtaining the final control point number T according to the frequency segment number N and the control point number M comprises:
when M is larger than or equal to N, only taking the first N segments of the M control points which are arranged in a descending order, namely the number T of the final control points is N;
and when M is less than N, the frequency bands capable of being divided in the vibration signals are less than the predefined frequency band number N, and the final control point number T is M.
5. The training method of the bearing fault diagnosis model according to claim 1, wherein screening a plurality of the modal signal components by kurtosis and correlation coefficients to obtain a reconstructed signal comprises:
calculating Kurt value of each modal signal componentnAnd retaining Kurt valuenAll components greater than 3 to form a set U;
calculating a correlation coefficient rho of each modal signal component and the fault vibration signal ynAnd with a correlation coefficient pnThe standard deviation of the mode signal is the lowest threshold value, and each mode signal component is screened to be used as the main rotation component of the fault vibration signal to form a set V;
and taking the union of the set U and the set V and adding signals therein to obtain a reconstructed signal sequence.
6. The training method of the bearing fault diagnosis model according to claim 5, wherein the kurtosis value Kurt isnBy the formula:
Figure FDA0003550397680000021
calculated to obtain wherein efdn,kK points for the nth modal signal component; n-1, 2, …, T; k is 1,2, …, L; u and σ are efdnMean and variance of; e [ n ]]Representing the mathematical expectation of n.
7. The bearing fault diagnosis method according to claim 5, wherein the correlation coefficient ρ isnBy the formula:
Figure FDA0003550397680000022
and calculating to obtain the result, wherein,
Figure FDA0003550397680000023
represents the mean value of the n-th modal component, zkThe kth sample point representing the fault vibration signal z,
Figure FDA0003550397680000024
representing the mean of all sample points of the fault vibration signal z.
8. The training method of the bearing fault diagnosis model according to claim 1, wherein the processing the reconstructed signal sequence by using the improved hierarchical slope entropy algorithm to obtain fault feature information and form a training feature set comprises:
given a reconstructed signal sequence x ═ xiI 1,2, …, L, and calculating the average operator P0And difference operator P1：
Figure FDA0003550397680000025
Wherein x isiA reconstructed signal representing an i-th section;
defining operators at different layer numbers s
Figure FDA0003550397680000026
Wherein,
Figure FDA0003550397680000031
given a particular sequence of vectors r1,r2,…,rs]And determining the f value of each layer according to the node e of the s-th layer, wherein the relationship between the node e and the vector sequence is as follows:
Figure FDA0003550397680000032
wherein, rdD ═ 1,2, …, s } - [0,1] denotes an average operator or a difference operator at the d-th layer;
the hierarchical components of the reconstructed signal sequence are obtained, expressed as:
Figure FDA0003550397680000033
calculating an improved level slope entropy for each level component, expressed as: MHSE (X, m, s, e, γ, δ) SE (X)s,eM, γ, δ) where m is the embedding dimension and γ and δ are the thresholds of the slope entropy.
9. The method for training the bearing fault diagnosis model according to claim 1, wherein the step of training the bearing fault diagnosis model by using the training feature set to obtain the trained bearing fault diagnosis model comprises:
the training feature set is denoted as { u }iI ═ 1,2, …, p }, and the weight between the input layer and the hidden layer is defined as
Figure FDA0003550397680000037
The weight between the hidden and output layers is β, the hidden layer threshold is b, p is 2s*(m-1)：
Figure FDA0003550397680000034
b＝[b1,b2,…,bQ]TR, Q, F represents the number of neurons in the input layer, hidden layer, and output layer of the extreme learning machine, respectively;
setting an activation function g (x), the output T of the extreme learning machine*Expressed as: t is*H β, wherein
Figure FDA0003550397680000035
Solving the minimum value of the approximate square difference to obtain the weight of the output layer as
Figure FDA0003550397680000036
And find its optimal solution as beta*＝H+T*In the formula, H+The generalized inverse of the output matrix for the hidden layer.
10. A bearing fault diagnosis method, comprising:
collecting a fault vibration signal of a bearing;
performing empirical Fourier decomposition on the fault vibration signal to obtain a plurality of modal signal components;
screening a plurality of modal signal components through kurtosis and correlation coefficients respectively, and reconstructing screening results to obtain a reconstructed signal sequence;
processing the reconstructed signal sequence by using an improved hierarchical slope entropy algorithm to obtain fault characteristic information;
inputting the fault characteristic information into a bearing fault diagnosis model trained by the bearing fault diagnosis model training method according to any one of claims 1 to 9 to obtain a diagnosis result of the bearing fault.
11. A bearing failure diagnosis device characterized by comprising:
the fault signal acquisition module is used for acquiring fault vibration signals of different types of faults of the bearing to obtain a sample set;
the modal signal acquisition module is connected with the fault signal acquisition module and is used for performing empirical Fourier decomposition on the fault vibration signals in the sample set to acquire a plurality of modal signal components;
the signal reconstruction module is connected with the modal signal acquisition module and is used for screening a plurality of modal signal components through kurtosis and correlation coefficients respectively and reconstructing a screening result to acquire a reconstructed signal sequence;
the training feature set acquisition module is connected with the signal reconstruction module and used for processing the reconstructed signal sequence by utilizing an improved hierarchical slope entropy algorithm to acquire fault feature information and form a training feature set;
the model construction module is connected with the training feature set acquisition module and used for training the bearing fault diagnosis model by using the training feature set so as to acquire a trained bearing fault diagnosis model;
and the fault diagnosis module is connected with the model construction module and is used for acquiring a diagnosis result.
CN202210259973.1A 2022-03-16 2022-03-16 Training method, diagnosis method and diagnosis device of bearing fault diagnosis model Pending CN114608826A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210259973.1A CN114608826A (en) 2022-03-16 2022-03-16 Training method, diagnosis method and diagnosis device of bearing fault diagnosis model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210259973.1A CN114608826A (en) 2022-03-16 2022-03-16 Training method, diagnosis method and diagnosis device of bearing fault diagnosis model

Publications (1)

Publication Number Publication Date
CN114608826A true CN114608826A (en) 2022-06-10

Family

ID=81862309

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210259973.1A Pending CN114608826A (en) 2022-03-16 2022-03-16 Training method, diagnosis method and diagnosis device of bearing fault diagnosis model

Country Status (1)

Country Link
CN (1) CN114608826A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115453336A (en) * 2022-08-25 2022-12-09 广东工业大学 GIS disconnecting link running state judging method and system and computer readable storage medium
CN115560966A (en) * 2022-09-30 2023-01-03 新黎明科技股份有限公司 Weak fault diagnosis method and system for key components of enhanced sparse filtering fusion motor
CN116165473A (en) * 2023-04-26 2023-05-26 广东工业大学 Real-time tracing method for network side overcurrent faults of train traction transmission system
CN116383635A (en) * 2023-06-05 2023-07-04 人工智能与数字经济广东省实验室(广州) Structure damage rapid identification method based on improved extreme learning machine
CN118025256A (en) * 2024-04-11 2024-05-14 大连新天勤轨道交通有限公司 On-line monitoring system for temperature of vehicle bearing

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20090080777A (en) * 2008-01-22 2009-07-27 성균관대학교산학협력단 Method and Apparatus for detecting signal
CN108168890A (en) * 2018-02-10 2018-06-15 北京工业大学 A kind of method of rolling bearing fault diagnosis
CN109100143A (en) * 2018-07-06 2018-12-28 华中科技大学 Fault Diagnosis of Roller Bearings and equipment based on CEEMDAN and CFSFDP
CN113780056A (en) * 2021-07-22 2021-12-10 山东科技大学 Rolling bearing fault diagnosis method based on vibration signal
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

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20090080777A (en) * 2008-01-22 2009-07-27 성균관대학교산학협력단 Method and Apparatus for detecting signal
CN108168890A (en) * 2018-02-10 2018-06-15 北京工业大学 A kind of method of rolling bearing fault diagnosis
CN109100143A (en) * 2018-07-06 2018-12-28 华中科技大学 Fault Diagnosis of Roller Bearings and equipment based on CEEMDAN and CFSFDP
CN113780056A (en) * 2021-07-22 2021-12-10 山东科技大学 Rolling bearing fault diagnosis method based on vibration signal
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

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115453336A (en) * 2022-08-25 2022-12-09 广东工业大学 GIS disconnecting link running state judging method and system and computer readable storage medium
CN115453336B (en) * 2022-08-25 2024-03-12 广东工业大学 GIS disconnecting link running state judging method, GIS disconnecting link running state judging system and computer readable storage medium
CN115560966A (en) * 2022-09-30 2023-01-03 新黎明科技股份有限公司 Weak fault diagnosis method and system for key components of enhanced sparse filtering fusion motor
CN115560966B (en) * 2022-09-30 2023-09-15 新黎明科技股份有限公司 Weak fault diagnosis method and system for key components of reinforced sparse filtering fusion motor
CN116165473A (en) * 2023-04-26 2023-05-26 广东工业大学 Real-time tracing method for network side overcurrent faults of train traction transmission system
CN116383635A (en) * 2023-06-05 2023-07-04 人工智能与数字经济广东省实验室(广州) Structure damage rapid identification method based on improved extreme learning machine
CN116383635B (en) * 2023-06-05 2024-04-09 人工智能与数字经济广东省实验室(广州) Structure damage rapid identification method based on improved extreme learning machine
CN118025256A (en) * 2024-04-11 2024-05-14 大连新天勤轨道交通有限公司 On-line monitoring system for temperature of vehicle bearing

Similar Documents

Publication Publication Date Title
CN114608826A (en) Training method, diagnosis method and diagnosis device of bearing fault diagnosis model
CN112101220B (en) Rolling bearing service life prediction method based on unsupervised model parameter migration
Wu et al. A hybrid classification autoencoder for semi-supervised fault diagnosis in rotating machinery
CN108344564B (en) A kind of state recognition of main shaft features Testbed and prediction technique based on deep learning
Shi et al. Rolling bearing initial fault detection using long short-term memory recurrent network
CN113344295B (en) Method, system and medium for predicting residual life of equipment based on industrial big data
CN111382542A (en) Road electromechanical equipment life prediction system oriented to full life cycle
Yan et al. Fisher’s discriminant ratio based health indicator for locating informative frequency bands for machine performance degradation assessment
Pan et al. A deep learning network via shunt-wound restricted Boltzmann machines using raw data for fault detection
CN115409131B (en) Production line abnormity detection method based on SPC process control system
CN113657454B (en) Nuclear power rotating machinery state monitoring method based on autoregressive BiGRU
Xu et al. A novel fault diagnosis method of bearing based on improved fuzzy ARTMAP and modified distance discriminant technique
Son et al. Deep learning-based anomaly detection to classify inaccurate data and damaged condition of a cable-stayed bridge
CN111538311A (en) Flexible multi-state self-adaptive early warning method and device for mechanical equipment based on data mining
Xin et al. An intelligent fault diagnosis method of rotating machinery based on deep neural networks and time-frequency analysis
CN110728377B (en) Intelligent fault diagnosis method and system for electromechanical equipment
Simula et al. Process monitoring and visualization using self-organizing maps
CN114861349A (en) Rolling bearing RUL prediction method based on model migration and wiener process
CN113505639B (en) Rotary machine multi-parameter health state assessment method based on TPE-XGBoost
Mahesh et al. Data-driven intelligent condition adaptation of feature extraction for bearing fault detection using deep responsible active learning
Techane et al. Rotating machinery prognostics and application of machine learning algorithms: Use of deep learning with similarity index measure for health status prediction
CN112132069A (en) Rolling bearing weak fault intelligent diagnosis method based on deep learning
CN114818779B (en) Fan rotor fault diagnosis method based on hierarchical classification algorithm
CN116541771A (en) Unbalanced sample bearing fault diagnosis method based on multi-scale feature fusion
CN114580472B (en) Large-scale equipment fault prediction method with repeated cause and effect and attention in industrial internet

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