[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

CN115015752A - Motor fault diagnosis method based on sparse decomposition and neighborhood bee colony algorithm - Google Patents

Motor fault diagnosis method based on sparse decomposition and neighborhood bee colony algorithm Download PDF

Info

Publication number
CN115015752A
CN115015752A CN202210618232.8A CN202210618232A CN115015752A CN 115015752 A CN115015752 A CN 115015752A CN 202210618232 A CN202210618232 A CN 202210618232A CN 115015752 A CN115015752 A CN 115015752A
Authority
CN
China
Prior art keywords
signal
motor
neighborhood
algorithm
search
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
CN202210618232.8A
Other languages
Chinese (zh)
Other versions
CN115015752B (en
Inventor
戴峻峰
常波
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.)
Huaiyin Institute of Technology
Original Assignee
Huaiyin Institute 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 Huaiyin Institute of Technology filed Critical Huaiyin Institute of Technology
Priority to CN202210618232.8A priority Critical patent/CN115015752B/en
Publication of CN115015752A publication Critical patent/CN115015752A/en
Application granted granted Critical
Publication of CN115015752B publication Critical patent/CN115015752B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/34—Testing dynamo-electric machines
    • G01R31/343—Testing dynamo-electric machines in operation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • 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
    • G06N3/08—Learning methods
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/80—Management or planning

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Control Of Electric Motors In General (AREA)

Abstract

The invention discloses a motor fault diagnosis method based on sparse decomposition and a neighborhood bee colony algorithm, which utilizes IM-HHT to extract direct current motor characteristics and analyze signals. Carrying out sparse decomposition denoising processing on a motor sampling signal, carrying out denoising before processing on the motor signal by using an orthogonal matching pursuit algorithm, and selecting a plurality of optimal atom linear combinations from an over-complete dictionary; decomposing a given signal into a plurality of IMFs by using empirical mode decomposition, removing false IMFs by using correlation parameters, and finally performing Hilbert transformation to obtain a Hilbert spectrum of the original signal; selecting characteristics based on a clustering Euclidean distance judgment method; sorting the motor signal characteristic factors by utilizing a square neighborhood selection artificial bee colony algorithm; and classifying and identifying the motor faults through a radial basis classifier. Compared with the prior art, the method can realize the automatic fault diagnosis of the motor aiming at the bearing fault, the stator winding fault and the rotor fault of the direct current motor.

Description

Motor fault diagnosis method based on sparse decomposition and neighborhood bee colony algorithm
Technical Field
The invention belongs to the technical field of motor fault diagnosis, and particularly relates to a motor fault diagnosis method based on sparse decomposition and a neighborhood swarm algorithm.
Background
The problem of motor faults commonly existing in manufacturing and production equipment is that at present, a general solution of a precision machining enterprise is as follows: on one hand, planned maintenance is carried out, namely, the motor is periodically overhauled and cleaned completely, the operation standard of production equipment is standardized as much as possible, the cultivation of operators and maintenance personnel is enhanced, the operators are familiar with the structure of the equipment, the performance of a motor used by the equipment is known, basic maintenance measures are understood, the maintenance and the maintenance of the equipment are enhanced, the normal operation of the production equipment is ensured, and the service life is prolonged. In addition, a large amount of manpower, material resources, financial resources and time are consumed for planning and overhauling, various resources are not reasonably utilized, the normal production time of an enterprise is reduced, the production efficiency is reduced, and the enterprise is actually caused with great economic loss. The traditional fault diagnosis methods have some defects, such as excessive dependence on prior knowledge, certain constraint on generalization capability and the like, and cannot meet the current requirement of motor fault monitoring.
Disclosure of Invention
The purpose of the invention is as follows: aiming at the problems in the prior art, the invention provides the motor fault diagnosis method based on sparse decomposition and the neighborhood bee colony algorithm, so that the system has higher precision and fewer characteristics, and the precision and the efficiency of fault diagnosis are improved.
The technical scheme is as follows: the invention provides a motor fault diagnosis method based on sparse decomposition and neighborhood swarm algorithm, comprising the following steps:
step 1: extracting the characteristics of the direct current motor and analyzing signals by using sparse decomposition IM-HHT; carrying out sparse decomposition denoising processing on a motor sampling signal, carrying out denoising before processing on the motor signal by using an orthogonal matching pursuit algorithm, selecting a plurality of optimal atoms from an overcomplete dictionary DCT, linearly combining the selected atoms, decomposing EMD (empirical mode decomposition), decomposing a given signal into a plurality of intrinsic mode functions IMF (intrinsic mode functions), removing false IMF by using correlation parameters, and finally carrying out Hilbert transformation to obtain a Hilbert spectrum of an original signal;
step 2: selecting characteristics based on a clustering Euclidean distance judgment method;
and step 3: selecting an artificial bee colony SS-ABC algorithm to sort the motor signal characteristic factors based on the square neighborhood; the square neighborhood selection artificial bee colony SS-ABC algorithm is a probability selection method which replaces an original ABC algorithm by using a square neighborhood selection method, a search strategy is modified and constructed, and improvement of the square neighborhood selection artificial bee colony SS-ABC algorithm is achieved;
and 4, step 4: and (4) classifying and identifying the motor faults through a radial basis RBF classifier.
Further, in the step 1, the sparse decomposition denoising process is performed, and the specific operation of denoising the motor signal before processing by using the orthogonal matching pursuit algorithm is as follows:
(1) characterizing signals to be decomposed and models
Figure BDA0003674388840000021
Wherein f is i For the signal to be decomposed, F ═ F 1 ，f 2 ，...f n D ═ D, m, and D ═ D 1 ，d 2 ，...d n Is an overcomplete dictionary, d i Is a selected atom, x i The sparse coefficient corresponding to the selected atom. The selected m atoms are linearly combined to be capable of sparsely representing signals;
finding x under the condition of satisfying formula (1) i I.e. seeking a solution with the fewest non-zero values, the model is as follows:
Figure BDA0003674388840000022
wherein | · | purple sweet 0 Is 1 0 Norm, i.e. x i Number of non-zero elements, X ═ X i The is a sparse coefficient matrix, and epsilon is an error;
(2) sparse decomposition
Selecting N-dimensional signal matrix F and atom D from overcomplete atom library D 1 Maximum value of inner product absolute value, d at this time 1 Is an optimal atom, namely satisfies the following conditions:
|<F，d 1 >|＝sup|<F，d i >| (3)
therein,. mu.g<F，d 1 >I represents the signal matrix F and the atom d 1 The absolute value of the inner product;
after the signal matrix F is decomposed for a plurality of times, the signal can be expressed as:
Figure BDA0003674388840000023
as the number of decompositions increases, the signal value of the residual signal will be approximately zero, and the signal matrix F to be decomposed may be expressed as:
Figure BDA0003674388840000024
further, the overcomplete dictionary DCT in step 1 is selected as follows:
a DCT dictionary is obtained from discrete cosine transform, and real numbers are used instead of complex numbers for analysis in the real number domain by symmetric signal extension, and the one-dimensional DCT transform can be calculated by the following formula:
Figure BDA0003674388840000031
wherein d (k) is the kth DCT atom, k is the frequency factor, f (N) is a given signal sequence, and N represents the input signal length; the matrix form is represented as follows:
D＝C N f (7)
C N the DCT coefficient matrix is adopted, and then the overcomplete dictionary can be obtained by performing fine sampling on the DCT coefficient matrix in a frequency domain.
Further, in the step 1, an empirical mode decomposition EMD is used to decompose a given signal into a plurality of intrinsic mode functions IMF, correlation parameters are used to remove false IMF, and finally Hilbert transform is performed to obtain a Hilbert spectrum of an original signal, which specifically includes:
1) decomposing a given signal into a plurality of intrinsic mode functions IMF by using Empirical Mode Decomposition (EMD);
let x (t) represent the original function of the input and be decomposed by EMD into n IMFs, the EMD model being defined as follows:
Figure BDA0003674388840000032
2) defining correlation parameters and removing false IMF
If the IMF component generated by EMD decomposition is x, and the motor fault signal is y, defining the correlation parameter as r:
Figure BDA0003674388840000033
wherein:
Figure BDA0003674388840000034
Figure BDA0003674388840000035
3) completing Hilbert transformation to obtain a corresponding Hilbert spectrum;
carrying out Hilbert transformation on the IMF of each motor state signal to obtain a Hilbert spectrum of the motor state signal; the HT model is defined as follows: :
Figure BDA0003674388840000036
where Pv is a warning value for avoiding an abnormality when τ ═ t and τ ± ∞ are given, the hilbert spectrum is defined as follows:
Figure BDA0003674388840000041
in the formula, h l (t) represents IMF, H l (t) can be obtained by Hilbert transform, a l (t) is the instantaneous amplitude, θ l (t) is the instantaneous phase angle.
Further, the specific operation of finding out the characteristic selection based on the clustering euclidean distance determining method in the step 2 includes:
1) calculate the variance of all samples of the mth feature
Figure BDA0003674388840000042
And average value
Figure BDA0003674388840000043
Figure BDA0003674388840000044
Wherein,
Figure BDA0003674388840000045
2) calculating the variance of the mth feature of the class C sample
Figure BDA0003674388840000046
And average value
Figure BDA0003674388840000047
Figure BDA0003674388840000048
Wherein,
Figure BDA0003674388840000049
3) calculating the clustering center g c Weighted variance at mth feature
Figure BDA00036743888400000410
Figure BDA00036743888400000411
Wherein,
Figure BDA00036743888400000412
4) calculating the distance d between classes of the mth feature b m And an intra-class distance d w m ：
Figure BDA00036743888400000413
Wherein,
Figure BDA00036743888400000414
5) in the mth feature, d is calculated b m Coefficient of variance v b m And d w m Of the variance factor v w m ：
Figure BDA0003674388840000051
Figure BDA0003674388840000052
Wherein,
Figure BDA0003674388840000053
6) calculating the compensation coefficient eta of the mth characteristic m ：
Figure BDA0003674388840000054
7) Calculating the distance discrimination factor lambda of the mth feature m And obtaining normalized lambda m′ ：
Figure BDA0003674388840000055
Figure BDA0003674388840000056
Further, the specific steps of selecting an artificial bee colony SS-ABC algorithm based on the square neighborhood to sort the motor signal characteristic factors in the step 3 are as follows:
1) employing bee search phase
Let X i ＝{X i，1 ，X i，2 ，...，X i，D Is the ith solution in the group, N is the total number of groups, D is the dimension, then the hiring bee at each solution X i (i 1, 2.., N) surrounding search, trying to find the optimal solution:
Figure BDA0003674388840000057
wherein x is k Randomly drawn from the entire population, jr is [1, D ]]Internal random integer, phi i，jr In [ -1, 1 [)]Randomly generating;
new solution v i The following are generated:
Figure BDA0003674388840000058
wherein j is 1, 2.. and D; a better solution is obtained using a greedy selection method, as follows:
Figure BDA0003674388840000059
2) following bee search phase
2.1) a square neighborhood selection method is used for replacing the original probability selection method, and based on the square neighborhood, a search strategy is modified and constructed:
the concept of a square neighborhood is utilized, a group solution is set to be a square topology, and whether a data point is in the neighborhood is judged by utilizing the following formula:
Figure BDA0003674388840000061
wherein x is i And y i Two square neighborhood points are provided, sl is the side length of a square neighborhood, and m is a neighborhood boundary coefficient;
2.2) modifying the search strategy: modifying and constructing a search strategy based on the side length of the square neighborhood;
for each solution x i Selecting x i As x, the best solution in the square neighborhood of ib Then using the corresponding search strategy at x ib Searching nearby; in the search process, x i Is x by ib Alternative, i.e. the follower bee will not search for x i But only x ib A neighborhood; the definition is as follows:
Figure BDA0003674388840000062
wherein x is ib Is from x i Is selected as the best solution in the square neighborhood, the weighting factor phi i，jr ∈[-1，+1]，
Figure BDA0003674388840000063
For the ith following bee, x ib Will be from x i The specific improved search method comprises the following steps:
Figure BDA0003674388840000064
wherein j is 1, 2.. D;
for v ib And x ib A better solution would be determined by:
Figure BDA0003674388840000065
3) searching stage for exploring bees
The new solution competes with its previous generation solution when v ib Is better than x ib Meaning in x ib The neighborhood search is successful; when v is ib Ratio x ib If the difference is not equal, the neighborhood search fails; failure by parameter real i The characterization is shown in the following formula:
Figure BDA0003674388840000066
if triali is greater than the limiting parameter, x ib I.e., discarded, a new solution for replacement will be created by:
x i，j ＝lw j +rd(0，1)*(u j -lw j ) (29)
wherein j is 1, 2, D, rd (0, 1) is [0, 1 ]]M random number, u j And lw j Is a boundary.
Further, the specific operation of classifying and identifying the motor fault through the radial basis RBF classifier in the step 4 is as follows: randomly selecting 70% of data samples from the motor sampling signal as training samples, the remaining 30% of data samples as test samples,
1) determining each initialization parameter: input vector X, actual output Y and expected output O, connection weight W (hidden layer-output layer), and center parameter C j (hidden layer), width vector D;
2) iteration of an algorithm
Figure BDA0003674388840000071
Figure BDA0003674388840000072
Figure BDA0003674388840000073
Wherein, W k，j (t) is the weight between the kth output neuron and the jth hidden neuron at the t iteration, C i，j (t) central component of the ith hidden layer neuron to the jth input neuron in the tth iteration, D i，j (t) is the width corresponding to the center, η is the learning factor, E is the evaluation function, defined as follows:
Figure BDA0003674388840000074
wherein o is l，k For the expected output of the kth output neuron at the l input sample, y l，k The output of the kth output neuron at the l-th input sample.
Has the beneficial effects that:
the invention discloses a motor fault diagnosis method based on sparse decomposition (IM-HHT) and a neighborhood swarm algorithm (SS-ABC). Firstly, extracting direct current motor characteristics and signal analysis by using IM-HHT, carrying out sparse decomposition denoising processing on motor sampling signals, carrying out denoising before processing on the motor signals by using an orthogonal matching pursuit algorithm, selecting a plurality of optimal atoms from an over-complete dictionary, linearly combining the selected atoms, decomposing EMD by using an empirical mode proposed by Huang, decomposing a given signal into a plurality of inherent modal functions, removing a false IMF by using correlation parameters, and carrying out Hilbert transformation to obtain a Hilbert spectrum of the original signal; then, selecting characteristics based on a clustering Euclidean distance judgment method; then, a square neighborhood selection method is used for replacing probability selection of an original ABC algorithm, a search strategy is modified and constructed, improvement of an artificial bee colony (SS-ABC) algorithm based on square neighborhood selection is achieved, and sequencing of motor signal characteristic factors is completed; and finally, classifying and identifying the motor fault through a Radial Basis Function (RBF) classifier. The invention realizes the automatic fault diagnosis of the motor, thereby realizing the fine diagnosis and the online monitoring of the motor fault. The method realizes early fault diagnosis and health state prediction of the motor, thereby ensuring that managers can find the fault reason of the motor in time, preventing the fault from further worsening, and ensuring that the maintenance time is reasonably arranged for the fault motor, thereby saving the maintenance fund, reducing the economic loss and having very important significance for improving the production efficiency and the economic benefit of enterprises.
In addition, the method provides a reference example for the practical application of motor fault diagnosis, particularly the application of the fault diagnosis method of sparse decomposition (IM-HHT) and neighborhood swarm algorithm (SS-ABC), and compared with other traditional diagnosis methods, the method can obtain higher motor fault classification precision.
Drawings
FIG. 1 shows the experimental design and the specific implementation process of the motor fault diagnosis method based on IM-HHT and SS-ABC;
FIG. 2 is a schematic view of a dimensional parameter of a motor according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a motor winding lead according to an embodiment of the present invention;
FIG. 4 is an overall flow of the IM-HHT algorithm of the present invention;
FIG. 5 is a sparse denoising process of the present invention;
FIG. 6 is a flow chart of the EMD algorithm execution of the present invention;
FIG. 7 is a schematic diagram of the IMF generated by empirical mode decomposition of the BLDC Hall signal of the present invention;
FIG. 8 is a characteristic distribution diagram of the IM-HHT of the present invention;
FIG. 9 is a flow chart of the characteristic selection implementation of the Euclidean distance clustering-based determination method of the present invention;
FIG. 10 is a flow chart of the present invention algorithm for selecting an artificial bee colony (SS-ABC) based on a square neighborhood;
fig. 11 is a selective supervised radial basis network topology of the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present invention is not limited thereby.
Aiming at the problem of motor faults commonly existing in manufacturing production equipment, the invention provides a motor fault diagnosis method based on sparse decomposition (IM-HHT) and a neighborhood swarm algorithm (SS-ABC). The method comprises the steps of conducting sparse decomposition denoising processing on motor sampling signals, conducting denoising before processing on the motor signals by using an orthogonal matching pursuit algorithm, selecting a plurality of optimal atoms from an over-complete dictionary, linearly combining the selected atoms, decomposing EMD by using an empirical mode proposed by Huang, decomposing given signals into a plurality of Intrinsic Mode Functions (IMF), removing false IMF by using correlation parameters, and finally conducting Hilbert transformation to obtain Hilbert spectrums of the original signals. Then, selecting characteristics based on a clustering Euclidean distance judgment method; next, a square neighborhood selection method is used for replacing probability selection of an original ABC algorithm, a search strategy is modified and constructed, improvement of an artificial bee colony (SS-ABC) algorithm based on square neighborhood selection is achieved, and sequencing of motor signal characteristic factors is completed; and finally, classifying and identifying the motor fault through a Radial Basis Function (RBF) classifier.
The embodiment of the invention aims at bearing faults, stator winding faults and rotor faults of the direct current motor and realizes the automatic fault diagnosis of the motor. Bearing fault, stator winding fault and rotor fault data and normal state data of a direct current motor are taken as bases.
The invention discloses a motor fault diagnosis method based on sparse decomposition (IM-HHT) and a neighborhood swarm algorithm (SS-ABC), the experimental design and the specific implementation process are shown in figure 1, and the method comprises the following steps:
step (1), BLDC (180W/3000RPM/DC 24V) motor signals are sampled, Matlab software is used for analysis, a fault classification system is established, and the BLDC (180W/3000RPM/DC 24V) is set to have the following three faults: bearing failure, stator winding failure, and rotor failure.
In the specific implementation, a servo motor generates a torque opposite to that of the brushless direct current motor, the torque is used as a load, the brushless direct current motor completes driving operation, then a data acquisition system (NI PXIe-1073) is used for acquiring Hall signals of the brushless direct current motor, the sampling frequency is 800Hz, the measuring time is 1000s, and further the measuring data of the Hall signals of the brushless direct current motor can be obtained. The motor size parameters and the motor winding lead wire diagrams are shown in figures 2 and 3.
And (2) extracting the characteristics of the direct current motor and analyzing signals by using IM-HHT. The method comprises the steps of conducting sparse decomposition denoising processing on motor sampling signals, conducting denoising before processing on the motor signals by using an orthogonal matching pursuit algorithm, selecting a plurality of optimal atoms from an over-complete dictionary, linearly combining the selected atoms, decomposing EMD by using an empirical mode proposed by Huang, decomposing given signals into a plurality of Intrinsic Mode Functions (IMFs), removing false IMFs by using correlation parameters, and finally conducting Hilbert transformation to obtain Hilbert spectrums of original signals.
In step (2):
in the process of extracting the characteristics of the Hall signals of the brushless direct current motors of four different types by utilizing an improved HHT algorithm (IM-HHT), firstly, after sparse decomposition and denoising processing is carried out on motor sampling signals, EMD is decomposed by utilizing an empirical mode proposed by Huang, a given signal is decomposed into a plurality of Intrinsic Mode Functions (IMF), then a pseudo IMF is removed by utilizing correlation parameters, and Hilbert transformation is carried out to obtain a Hilbert spectrum of an original signal.
The overall flow of the IM-HHT algorithm is shown in figure 4.
Step (2) a: sparse decomposition denoising
(1) Characterizing signals to be decomposed and models
The signal to be decomposed is represented by the following formula (1):
Figure BDA0003674388840000101
wherein f is i In order for the signal to be decomposed,F＝{f 1 ，f 2 ，...f n d ═ D, m, and D ═ D 1 ，d 2 ，...d n Is an overcomplete dictionary, d i Is a selected atom, x i The sparse coefficient corresponding to the selected atom. And the selected m atoms are linearly combined to sparsely represent the signal.
The core problem of sparse representation of signals is described as finding x under the condition of satisfying formula (1) i The most sparse representation of (a) is the value that the solution with the least non-zero value is sought, and the model is as shown in equation (2):
Figure BDA0003674388840000102
wherein | · | purple sweet 0 Is 1 0 Norm, i.e. x i Number of non-zero elements, X ═ X i Is the sparse coefficient matrix and epsilon is the error.
(2) Sparse decomposition
Selecting N-dimensional signal matrix F and atom D from overcomplete atom library D 1 Maximum value of inner product absolute value, d at this time 1 Is an optimum atom, namely, satisfy: ( 3 ) Formula (II):
|<F，d 1 >|＝sup|<F，d i 〉| (3)
wherein a component O is<F，d 1 Signal matrix F and atom d 1 Absolute value of inner product, d 1 Is the first matched optimal atom.
At this time, F can be expressed as:
F＝<F，d 1 >d 1 +R 1 F (4)
wherein,<F，d 1 >d 1 represents the signal matrix F at d 1 Projection of (a) onto, R 1 F denotes the remaining part of the signal matrix F after the first decomposition.
Thus, the k-th decomposed signal can be expressed as:
R k F＝<R k F，d k )d k +R k+1 F (5)
after the signal matrix F is decomposed for a plurality of times, the signal can be expressed as:
Figure BDA0003674388840000103
as the number of decompositions increases, the signal value of the residual signal will be approximately zero, and the signal matrix F to be decomposed may be expressed as:
Figure BDA0003674388840000111
the sparse denoising process is shown in fig. 5.
Note that: in the loop iteration process, the selected atoms and the residual signals meet the orthogonality, and the selected atoms are removed and are not used any more.
(3) Selection of DCT dictionary
In the implementation, a DCT dictionary is obtained according to discrete cosine transform, and analysis is carried out on a real number domain by replacing a complex number with a real number through symmetrical signal extension. The one-dimensional DCT transform can be calculated by:
Figure BDA0003674388840000112
Figure BDA0003674388840000113
where J (k) is the kth DCT atom, k is the frequency factor, f (N) is the given signal sequence, and N represents the input signal length. The matrix form is represented as follows:
D＝C N f (10)
C N the DCT coefficient matrix is adopted, and then the overcomplete dictionary can be obtained by performing fine sampling on the DCT coefficient matrix in a frequency domain.
Step (2) b: empirical Mode Decomposition (EMD)
A given signal is decomposed into several Intrinsic Mode Functions (IMFs), which are components that satisfy certain conditions, by Empirical Mode Decomposition (EMD) proposed by Huang.
Let x (t) denote the original function of the input and be decomposed by EMD into n IMFs, the EMD model being defined by the equation (11):
Figure BDA0003674388840000114
the EMD algorithm executes the flow chart shown in FIG. 6.
Step (2) c: removing spurious IMF using correlation parameters
In practice, the correlation parameter formula is defined as follows:
and if the IMF component generated by EMD decomposition is x and the motor fault signal is y, defining the correlation parameter as r:
Figure BDA0003674388840000115
wherein,
Figure BDA0003674388840000121
Figure BDA0003674388840000122
relevance parameter meaning:
when r is more than 0, the positive correlation of the two variables is shown, and when r is less than 0, the negative correlation of the two variables is shown;
when the | r | ═ 1, the two variables are completely linearly related, namely, the functional relationship is obtained;
when r is 0, the wireless correlation relationship between the two variables is represented;
when 0 < | r | < 1, it means that there is some linear correlation between the two variables. The closer the | r | is to 1, the more closely the linear relationship between the two variables is; the closer | r | is to 0, the weaker the linear correlation of the two variables.
Generally, the method can be divided into three stages: a low degree line of | < 0.4 | r |)Correlation is carried out; the absolute r is more than or equal to 0.4 and less than 0.7 is significance correlation; the linear correlation is higher when the absolute value of r is more than or equal to 0.7 and less than 1. In practice, the threshold value is set to be 0.5, if r i If the IMF component is more than 0.5, the IMF component is reserved; if r i If the IMF component is less than or equal to 0.5, the IMF component is determined to be false, and the IMF component is removed.
Step (2) d: hilbert transformation is carried out to obtain a corresponding Hilbert spectrum
Empirical Mode Decomposition (EMD) is carried out, and after removing false IMF by using the correlation parameter, a natural mode function (IMF) is introduced into Hilbert Transform (HT), so that the instantaneous amplitude and the instantaneous frequency of the signal are obtained. I.e., Hilbert transform is performed on the IMF of each motor state signal to obtain a Hilbert spectrum of the motor state signal. The HT model is defined as in formula (13):
Figure BDA0003674388840000123
where Pv is a warning value for avoiding abnormalities when τ ═ t and τ ± ∞ then the hilbert spectrum is defined as follows:
Figure BDA0003674388840000124
in the formula, h l (t) represents IMF, H l (t) can be obtained by Hilbert transform, a l (t) is the instantaneous amplitude, θ l (t) is the instantaneous phase angle.
In the specific implementation, BLDC (180W/3000RPM/DC 24V) motor signals are sampled, Matlab software is used for analysis, a fault classification system is established, and the following three faults exist in the BLDC (180W/3000RPM/DC 24V): bearing failure, stator winding failure, and rotor failure. The fault classification system was investigated for its ability to classify three different types of faults. After IM-HHT feature extraction and analysis are carried out on the measured sampling data, the extracted features capable of reflecting the motor state are normalized, and the feature value of the motor type is ensured to be between 0 and 1, so that the problem of gradient explosion in a classifier is avoided.
In the specific implementation, a torque opposite to that of the brushless direct current motor is generated by the servo motor and is used as a load, the brushless direct current motor completes driving operation, then a data acquisition system (NI PXIe-1073) is used for acquiring Hall signals of the brushless direct current motor, the sampling frequency is 800Hz, the measuring time is 1000s, and further the measuring data of the Hall signals of the brushless direct current motor can be obtained to research the fault state and the performance of the direct current motor. A total of four BLDCs were tested in the experiment, one motor was normal and the other three had a fault, the brushless dc motor used was rated at 24V, the motor rated speed was configured at 3000RPM, and the parameters of the brushless dc motor are listed in table 1.
TABLE 1 BLDC parameters
Type (B) Rated current Rated value Rated speed of rotation Rated output power Rated efficiency
57BL115S18 22A 12Kg-cm 3000RPM 180W 83％
Four types of motor hall signals are processed through EMD decomposition in an improved HHT algorithm (IM-HHT), the signals are decomposed into a first layer to a sixth layer (IMF1 to IMF6), and instantaneous amplitude and instantaneous frequency of each layer are obtained through Hilbert-Huang transformation. In addition, in the time domain, the maximum value (Tmax), the minimum value (Tmin), the average value (Tmean), the mean square error (Tme), and the standard deviation (Tstd) are captured. In the frequency domain, the maximum value (Fmax), the average value (Fmean), the mean square error (Fme), and the standard value deviation (Fstd) are obtained. Each IMF extracts 10 features and normalizes them so that their feature values are distributed between 0 and 1. A total of 60 features were obtained as shown in table 2.
TABLE 2 HHT extracted Motor Signal characteristic parameters
Figure BDA0003674388840000131
In implementation, the IMF waveform generated by empirical mode decomposition of the BLDC Hall signal is shown in FIG. 7. The transform first extracts a signal high frequency, the subsequent layers of IMF are low frequency waveforms, and fig. 8 is a characteristic distribution plot of HHT.
And (3) selecting characteristics based on a clustering Euclidean distance judgment method.
The characteristic selection based on the clustering Euclidean distance judgment method is carried out, a characteristic selection stage is used for selecting a few key characteristics, in the implementation process, the separability of characteristic categories is calculated by using the clustering Euclidean distance judgment based method, wherein high calculation factors represent important characteristics. Characteristic distance discrimination factor lambda m Are based on the same category
Figure BDA0003674388840000141
And different classes of features of
Figure BDA0003674388840000142
The euclidean distance between them. The Euclidean distance of a feature is determined by the center of the class feature
Figure BDA0003674388840000143
And sample characteristicsCenter (C)
Figure BDA0003674388840000144
And (6) calculating. Wherein c, m, i are class number, feature number and sample number,
Figure BDA0003674388840000145
is a characteristic of the sample. Compensation factor eta m By distance variance
Figure BDA0003674388840000146
And
Figure BDA0003674388840000147
the calculation is obtained and the calculation process is as follows:
step (3) a: calculate the variance of all samples of the mth feature
Figure BDA0003674388840000148
And average value
Figure BDA0003674388840000149
Figure BDA00036743888400001410
Wherein,
Figure BDA00036743888400001411
step (3) b: calculating the variance of the mth feature of the class C sample
Figure BDA00036743888400001412
And average value
Figure BDA00036743888400001413
Figure BDA00036743888400001414
Wherein,
Figure BDA00036743888400001415
step (3) c: calculating a clustering center g c Weighted variance at mth feature
Figure BDA00036743888400001416
Figure BDA00036743888400001417
Wherein,
Figure BDA00036743888400001418
step (3) d: calculating the distance d between the m-th features b m And an intra-class distance d w m
Figure BDA00036743888400001419
Wherein,
Figure BDA0003674388840000151
step (3) e: in the mth feature, d is calculated b m Coefficient of variance of (1) vbm and d w m Variance factor vwm of
Figure BDA0003674388840000152
Wherein,
Figure BDA0003674388840000153
step (3) g: calculating the compensation coefficient eta of the mth characteristic m And distance discrimination factor lambda m And normalized lambda m′
Figure BDA0003674388840000154
A feature selection flow chart of the clustering-based euclidean distance determination method is shown in fig. 9.
And (4) replacing the probability selection of the original ABC algorithm with a square neighborhood selection method, modifying and constructing a search strategy, realizing the improvement of selecting an artificial bee colony (SS-ABC) algorithm based on the square neighborhood, and finishing the sequencing of the motor signal characteristic factors.
After the characteristics are selected, the selected characteristic factors are arranged in a descending order by utilizing an improved Artificial Bee Colony (ABC) algorithm, and because the characteristics of some Hall signals can reduce the recognition rate of the classifier or the recognition result is not influenced, the characteristics can be deleted through the characteristic selection, so that the calculation time is saved. And the calculation efficiency or the recognition rate can be improved by combining with an Artificial Bee Colony (ABC) algorithm. The invention sets the recognition result as the fitness value, inputs the fitness value as the characteristic grade, outputs the characteristic grade as the new characteristic grade, sequences the characteristics, sequences the optimized characteristics through the ABC algorithm, and can improve the performance and the identity of the characteristics.
An Artificial Bee Colony (ABC) algorithm is a powerful tool problem for solving an optimization problem, belongs to a branch of artificial intelligence, and simulates the behavior of a bee colony in the process of developing and discarding a food source searched by the ABC algorithm. The location of the food source indicates a possible way of solving the problem, and in order to find the best solution, the artificial bees interact and exchange information, concentrate on the desired large solution interval, and steadily leave the desired less solution area by using collective knowledge, whereby the solution is collectively improved step by step, and the search is repeated by the algorithm until a predetermined stop condition is met.
In the ABC algorithm, bees are divided into three different groups: hire bees, follow bees, reconnaissance bees. Employing bees to master initial honey source information and sending the information to the circled bees through swinging dance; the follower bees select to follow the honey collection according to the judgment of the follower bees, the reconnaissance bees judge whether to give up the old honey source information or not to search for a new honey source for replacement, the follower bees search the surrounding environment, the new honey source is stored in the memory every time the new honey source is found, and if the new honey source is better than the previous honey source, the bee remembers the new situation and forgets the previous situation. The three stages are continuously and repeatedly executed until an optimal solution is obtained.
Step (4) a: search stage of hiring bees
Let X i ＝{X i，1 ，X i，2 ，...，X i，D Is the ith solution in the group, N is the total number of groups, D is the dimension, then the hiring bee at each solution X i (i 1, 2.., N) surrounding search, trying to find the optimal solution, see formula (21)
Figure BDA0003674388840000161
Wherein x is k Randomly drawn from the whole population, jr is [1, D ]]Internal random integer, phi i，jr In [ -1, 1 [)]Randomly generated, whereby a new solution v i Generating the formula (22):
Figure BDA0003674388840000162
wherein j is 1, 2.. and D;
next, a better solution is obtained using a greedy selection method, as shown in equation (23).
Figure BDA0003674388840000163
Step (4) b: following bee search phase
At this stage, the hiring bee completes a neighborhood search of all solutions, and the following bee obtains search information from the hiring bee. Unlike the employment bees, in this stage, the following bees select only part of the better solutions to further search. In the invention, an improved ABC algorithm method based on neighborhood selection is provided, namely a square neighborhood selection artificial bee colony (SS-ABC). The main improvement lies in that the original probability selection method is replaced by a square neighborhood selection method, and a search strategy is modified and constructed on the basis of the square neighborhood.
(1) Square neighborhood selection
And setting the group solution as a square topology by using the concept of square neighborhood. And judging whether the data point is in the neighborhood by the formula (24):
Figure BDA0003674388840000164
wherein x is i And y i Is two square neighborhood points, sl is the square neighborhood side length, and m is the neighborhood boundary coefficient.
When x is i And y i The method satisfies the formula (24), namely the data points in the neighborhood are determined, and the algorithm does not need multiplication operation and evolution operation, so that the clustering efficiency can be improved.
(2) Modifying search strategies
And modifying and constructing a search strategy based on the side length of the square neighborhood. A follower bee search scheme selection mechanism is proposed, for each solution x i Selecting x i As x, the best solution in the square neighborhood of ib Then using the corresponding search strategy at x ib And (5) searching nearby. Compared with the traditional ABC algorithm, the method does not need to calculate the selection probability of each solution.
In the search process, x i Is x by ib Alternative, i.e. follower bee will not search for x i To search only x ib A neighborhood. As defined in formula (25)
Figure BDA0003674388840000171
Wherein x is ib Is from x i Is selected as the best solution in the square neighborhood, the weighting factor phi i，jr ∈[-1，+1]，
Figure BDA0003674388840000172
For the ith following bee, x ib Will be from x i In a square neighborhood of the search, in particular refining the searchThe method comprises the following steps:
Figure BDA0003674388840000173
Figure BDA0003674388840000174
wherein j is 1, 2.
For v ib And x ib A better solution will be determined by equation (28):
Figure BDA0003674388840000175
step (4) c: searching stage for exploring bees
The new solution competes with its previous generation. When v is i Is better than x i Meaning at x i The neighborhood search is successful. When v is i Ratio x i Poor, it means that the neighborhood search failed. Failure count by parameter real i Characterisation, when the search is successful, the real i 0, when the search fails, the real i Added one, described in detail in equation (29):
Figure BDA0003674388840000176
if the deal i Greater than the limiting parameter, corresponding to a solution x i I.e., discarded, a new solution for replacement will be created by equation (30):
x i，j ＝lw j +rd(0，1)*(u j -lw j ) (30)
wherein j is 1, 2, D, rd (0, 1) is [0, 1 ]]M random number, u j And lw j Is a boundary.
The algorithm flow for selecting artificial bee colony (SS-ABC) based on square neighborhood is shown in fig. 10.
And (5) classifying and identifying the motor faults by using a Radial Basis Function (RBF) classifier.
Building Radial Basis Function (RBF) classifiers
And (4) classifying and identifying the motor fault through a Radial Basis Function (RBF) classifier. From the motor sample signal, 70% of the data samples were randomly selected as training samples, and the remaining 30% of the data were used as test samples. A Radial Basis Function (RBF) network is generally a three-layer structure, and a topology structure of a selective supervision type RBF network is shown in fig. 11.
The nodes of the input layer and the output layer of the radial basis function neural network are linear functions, and the hidden layer is a radial basis function and has local approximation capability. In the iteration process of the RBF algorithm, a gradient descent method is adopted for self-adaptively adjusting the center, the width and the weight, and the specific steps are as follows:
step (5) a: determining an input vector X
X＝{x 1 ，x 2 ，…x n } T (31)
Wherein n is the number of input neurons.
Step (5) b: determining the actual output Y and the desired output O
Y＝{y 1 ，y 2 ，…y q } T ，O＝{o 1 ，o 2 ，…o q } T (32)
Wherein q is the number of output neurons.
Step (5) c: initialization connection weight W (hidden layer-output layer)
W k ＝{w k，1 ，w k，2 ，…w k，p } T (33)
Wherein p is the number of cryptic neurons.
Step (5) d: initialization center parameter C j (hidden layer)
C j ＝{c j，1 ，c j，2 ，…c j，n } T (34)
Step (5) e: initializing width vector D
D i ＝{d i，1 ，d i，2 ，…d i，n } T
Figure BDA0003674388840000181
Wherein d is an adjustment coefficient
Step (5) f: the algorithm iterates until an error condition is met or a maximum number of iterations is reached
Figure BDA0003674388840000182
Figure BDA0003674388840000183
Figure BDA0003674388840000191
Wherein, W k，j (t) is the weight between the kth output neuron and the jth hidden neuron at the t iteration, C i，j (t) central component of the ith hidden layer neuron to the jth input neuron in the tth iteration, D i，j (t) is the width corresponding to the center, η is the learning factor, E is the evaluation function, defined as follows:
Figure BDA0003674388840000192
wherein o is l，k For the expected output of the kth output neuron at the l input sample, y l，k The output of the kth output neuron at the l-th input sample.
The above embodiments are merely illustrative of the technical concepts and features of the present invention, and the purpose of the embodiments is to enable those skilled in the art to understand the contents of the present invention and implement the present invention, and not to limit the protection scope of the present invention. All equivalent changes and modifications made according to the spirit of the present invention should be covered in the protection scope of the present invention.

Claims (7)

1. A motor fault diagnosis method based on sparse decomposition and neighborhood bee colony algorithm is characterized by comprising the following steps:
step 1: extracting the characteristics of the direct current motor and analyzing signals by using sparse decomposition IM-HHT; carrying out sparse decomposition denoising processing on a motor sampling signal, carrying out denoising before processing on the motor signal by using an orthogonal matching pursuit algorithm, selecting a plurality of optimal atoms from an overcomplete dictionary DCT, linearly combining the selected atoms, decomposing EMD (empirical mode decomposition), decomposing a given signal into a plurality of intrinsic mode functions IMF (intrinsic mode functions), removing false IMF by using correlation parameters, and finally carrying out Hilbert transformation to obtain a Hilbert spectrum of an original signal;
step 2: selecting characteristics based on a clustering Euclidean distance judgment method;
and step 3: selecting an artificial bee colony SS-ABC algorithm to sort the motor signal characteristic factors based on the square neighborhood; the square neighborhood selection artificial bee colony SS-ABC algorithm is a probability selection method which replaces an original ABC algorithm by using a square neighborhood selection method, a search strategy is modified and constructed, and improvement of the square neighborhood selection artificial bee colony SS-ABC algorithm is achieved;
and 4, step 4: and (4) classifying and identifying the motor faults through a radial basis RBF classifier.
2. The motor fault diagnosis method based on sparse decomposition and neighborhood swarm algorithm as claimed in claim 1, wherein the sparse decomposition denoising processing in step 1, and the specific operation of denoising the motor signal before processing by using the orthogonal matching pursuit algorithm, is as follows:
(1) characterizing signals to be decomposed and models
Figure FDA0003674388830000011
Wherein f is i For the signal to be decomposed, F ═ F 1 ，f 2 ，...f n Is the signal matrix to be decomposed, m is the selected primitiveNumber of children, D ═ D 1 ，d 2 ，...d n Is an overcomplete dictionary, d i Is a selected atom, x i For the sparse coefficient corresponding to the selected atoms, performing linear combination on the selected m atoms to obtain sparse representation of the signal;
finding x under the condition of satisfying formula (1) i I.e. seeking a solution with the fewest non-zero values, the model is as follows:
Figure FDA0003674388830000012
wherein | · | purple sweet 0 Is 1 of 0 Norm, i.e. x i Number of non-zero elements, X ═ X i } is the sparse coefficient matrix, ε is the error;
(2) sparse decomposition
Selecting N-dimensional signal matrix F and atom D from overcomplete atom library D 1 Maximum value of inner product absolute value, d at this time 1 Is an optimal atom, namely satisfies the following conditions:
|<F，d 1 >|＝sup|<F，d i >| (3)
therein,. mu.g<F，d 1 >I represents the signal matrix F and the atom d 1 The absolute value of the inner product;
after the signal matrix F is decomposed for a plurality of times, the signal can be expressed as:
Figure FDA0003674388830000021
as the number of decompositions increases, the signal value of the residual signal will be approximately zero, and the signal matrix F to be decomposed may be expressed as:
Figure FDA0003674388830000022
3. the motor fault diagnosis method based on sparse decomposition and neighborhood bee colony algorithm according to claim 1, wherein the selection of the overcomplete dictionary DCT in the step 1 is as follows:
a DCT dictionary is obtained from discrete cosine transform, and a real number instead of a complex number is analyzed in a real number domain by symmetric signal extension, and the one-dimensional DCT transform can be calculated by:
Figure FDA0003674388830000023
wherein d (k) is the kth DCT atom, k is the frequency factor, f (N) is a given signal sequence, and N represents the input signal length; the matrix form is represented as follows:
D＝C N f (7)
C N the DCT coefficient matrix is adopted, and then the overcomplete dictionary can be obtained by performing fine sampling on the DCT coefficient matrix in a frequency domain.
4. The motor fault diagnosis method based on sparse decomposition and neighborhood swarm algorithm according to claim 1, wherein the step 1 is to decompose a given signal into a plurality of intrinsic mode functions IMFs by using empirical mode decomposition EMD, remove spurious IMFs by using correlation parameters, and finally perform Hilbert transform to obtain the Hilbert spectrum of the original signal, which specifically comprises the following operations:
1) decomposing a given signal into a plurality of intrinsic mode functions IMF by using Empirical Mode Decomposition (EMD);
let x (t) represent the original function of the input and be decomposed by EMD into n IMFs, the EMD model being defined as follows:
Figure FDA0003674388830000031
2) defining correlation parameters and removing false IMF
And if the IMF component generated by EMD decomposition is x and the motor fault signal is y, defining the correlation parameter as r:
Figure FDA0003674388830000032
wherein:
Figure FDA0003674388830000033
Figure FDA0003674388830000034
3) completing Hilbert transformation to obtain a corresponding Hilbert spectrum;
carrying out Hilbert transformation on the IMF of each motor state signal to obtain a Hilbert spectrum of the motor state signal; the HT model is defined as follows: :
Figure FDA0003674388830000035
where Pv is a warning value for avoiding abnormalities when τ ═ t and τ ± ∞ then the hilbert spectrum is defined as follows:
Figure FDA0003674388830000036
in the formula, h l (t) represents IMF, H l (t) can be obtained by Hilbert transform, a l (t) is the instantaneous amplitude, θ l (t) is the instantaneous phase angle.
5. The motor fault diagnosis method based on sparse decomposition and neighborhood bee colony algorithm according to claim 1, wherein the specific operation of finding out the characteristic selection based on the clustering Euclidean distance judgment method in the step 2 comprises the following steps:
1) calculate the variance of all samples of the mth feature
Figure FDA0003674388830000037
And average value
Figure FDA0003674388830000038
Figure FDA0003674388830000041
Wherein,
Figure FDA0003674388830000042
2) calculating the variance of the mth feature of the class C sample
Figure FDA0003674388830000043
And average value
Figure FDA0003674388830000044
Figure FDA0003674388830000045
Wherein,
Figure FDA0003674388830000046
3) calculating the clustering center g c Weighted variance at mth feature
Figure FDA0003674388830000047
Figure FDA0003674388830000048
Wherein,
Figure FDA0003674388830000049
4) calculating the distance d between the m-th features b m And an intra-class distance d w m ：
Figure FDA00036743888300000410
Wherein,
Figure FDA00036743888300000411
5) in the mth feature, d is calculated b m Coefficient of variance v b m And d w m Of the variance factor v w m ：
Figure FDA00036743888300000412
Figure FDA00036743888300000413
Wherein,
Figure FDA00036743888300000414
6) calculating the compensation coefficient eta of the mth characteristic m ：
Figure FDA0003674388830000051
7) Calculating the distance discrimination factor lambda of the mth feature m And obtaining normalized lambda m′ ：
Figure FDA0003674388830000052
Figure FDA0003674388830000053
。
6. The motor fault diagnosis method based on sparse decomposition and neighborhood bee colony algorithm of claim 1, wherein the step 3 of selecting the artificial bee colony SS-ABC algorithm based on the square neighborhood to sequence the motor signal characteristic factors comprises the following specific steps:
1) employing bee search phase
Let X i ＝{X i，1 ，X i，2 ，...，X i，D Is the ith solution in the group, N is the total group number, D is the dimension, then hire bees at each solution X i (i 1, 2.., N) surrounding search, trying to find the optimal solution:
Figure FDA0003674388830000054
wherein x is k Randomly drawn from the whole population, jr is [1, D ]]Internal random integer, phi i，jr In [ -1, 1 [)]Randomly generating;
new solution v i The following are generated:
Figure FDA0003674388830000055
wherein j is 1, 2.. and D; a better solution is obtained using a greedy selection method, as follows:
Figure FDA0003674388830000056
2) following bee search phase
2.1) a square neighborhood selection method is used for replacing the original probability selection method, and based on the square neighborhood, a search strategy is modified and constructed:
the concept of square neighborhood is utilized, the group solution is set to be square topology, and whether the data point is in the neighborhood is judged by the following formula:
Figure FDA0003674388830000061
wherein x is i And y i Two square neighborhood points are provided, sl is the side length of a square neighborhood, and m is a neighborhood boundary coefficient;
2.2) modifying the search strategy: modifying and constructing a search strategy based on the side length of the square neighborhood;
for each solution x i Selecting x i As x, the best solution in the square neighborhood of ib Then using the corresponding search strategy at x ib Searching nearby; in the search process, x i Is x by ib Alternative, i.e. the follower bee will not search for x i But only x ib A neighborhood; the definition is as follows:
Figure FDA0003674388830000062
wherein x is ib Is from x i Is selected as the best solution in the square neighborhood, the weighting factor phi i，jr ∈[-1，+1]，
Figure FDA0003674388830000063
For the ith following bee, x ib Will be from x i The specific improved search method comprises the following steps:
Figure FDA0003674388830000064
wherein j is 1, 2.. and D;
for v ib And x ib A better solution would be determined by:
Figure FDA0003674388830000065
3) searching stage for exploring bees
The new solution competes with its previous generation solution when v ib Is better than x ib Meaning at x ib The neighborhood search is successful; when v is ib Ratio x ib If the difference is not equal, the neighborhood search fails; failure by parameter real i The characterization is shown in the following formula:
Figure FDA0003674388830000066
if the real i Greater than a limiting parameter, x ib I.e., discarded, a new solution for replacement will be created by:
x i，j ＝lw j +rd(0，1)*(u j -lw j ) (29) wherein j is 1, 2, and D, rd (0, 1) is [0, 1 ]]M random number, u j And lw j Is a boundary.
7. The motor fault diagnosis method based on sparse decomposition and neighborhood bee colony algorithm of claim 1, wherein the specific operation of classifying and identifying the motor fault through the radial basis RBF classifier in the step 4 is as follows: randomly selecting 70% of data samples from the motor sampling signal as training samples, the remaining 30% of data as test samples,
1) determining each initialization parameter: input vector X, actual output Y and expected output O, connection weight W (hidden layer-output layer), center parameter C j (hidden layer), width vector D;
2) iteration of an algorithm
Figure FDA0003674388830000071
Figure FDA0003674388830000072
Figure FDA0003674388830000073
Wherein, W k，j (t) is the weight between the kth output neuron and the jth hidden neuron at the t iteration, C i，j (t) central component of the ith hidden layer neuron to the jth input neuron in the tth iteration, D i，j (t) is the width corresponding to the center, η is the learning factor, E is the evaluation function, defined as follows:
Figure FDA0003674388830000074
wherein o is l，k For the expected output of the kth output neuron at the l input sample, y l，k The output of the kth output neuron at the l-th input sample.
CN202210618232.8A 2022-06-01 2022-06-01 Motor fault diagnosis method based on sparse decomposition and neighborhood swarm algorithm Active CN115015752B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210618232.8A CN115015752B (en) 2022-06-01 2022-06-01 Motor fault diagnosis method based on sparse decomposition and neighborhood swarm algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210618232.8A CN115015752B (en) 2022-06-01 2022-06-01 Motor fault diagnosis method based on sparse decomposition and neighborhood swarm algorithm

Publications (2)

Publication Number Publication Date
CN115015752A true CN115015752A (en) 2022-09-06
CN115015752B CN115015752B (en) 2024-08-20

Family

ID=83073669

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210618232.8A Active CN115015752B (en) 2022-06-01 2022-06-01 Motor fault diagnosis method based on sparse decomposition and neighborhood swarm algorithm

Country Status (1)

Country Link
CN (1) CN115015752B (en)

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104089774A (en) * 2014-07-25 2014-10-08 北京工业大学 Gear fault diagnosis method based on orthogonal match between multiple parallel dictionaries
CN106441871A (en) * 2016-10-20 2017-02-22 哈尔滨工业大学 Wind power gearbox fault diagnosis method based on self-adaptive resonance sparse decomposition theory
CN110232699A (en) * 2019-05-05 2019-09-13 淮阴工学院 A kind of image multilayer feature decomposition method based on the sparse statistical property of L0
CN110398364A (en) * 2019-07-05 2019-11-01 东南大学 Epicyclic gearbox method for diagnosing faults based on resonance sparse decomposition and FastICA algorithm
CN110824293A (en) * 2019-10-15 2020-02-21 淮阴工学院 Power grid fault diagnosis method based on multi-feature fusion parameters of wolf pack algorithm
CN111665050A (en) * 2020-06-04 2020-09-15 燕山大学 Rolling bearing fault diagnosis method based on clustering K-SVD algorithm
WO2021074248A1 (en) * 2019-10-14 2021-04-22 Norwegian University Of Science And Technology (Ntnu) Fault detection in synchronous machines
CN113203565A (en) * 2021-03-25 2021-08-03 长江大学 Bearing fault identification method and system based on EEMD sparse decomposition
CN113310684A (en) * 2021-04-20 2021-08-27 东南大学 Gearbox fault feature extraction method based on scale space and improved sparse representation
US20210270244A1 (en) * 2020-03-02 2021-09-02 Wuhan University Method and system for fault diagnosis of gearbox of wind turbine generator

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104089774A (en) * 2014-07-25 2014-10-08 北京工业大学 Gear fault diagnosis method based on orthogonal match between multiple parallel dictionaries
CN106441871A (en) * 2016-10-20 2017-02-22 哈尔滨工业大学 Wind power gearbox fault diagnosis method based on self-adaptive resonance sparse decomposition theory
CN110232699A (en) * 2019-05-05 2019-09-13 淮阴工学院 A kind of image multilayer feature decomposition method based on the sparse statistical property of L0
CN110398364A (en) * 2019-07-05 2019-11-01 东南大学 Epicyclic gearbox method for diagnosing faults based on resonance sparse decomposition and FastICA algorithm
WO2021074248A1 (en) * 2019-10-14 2021-04-22 Norwegian University Of Science And Technology (Ntnu) Fault detection in synchronous machines
CN110824293A (en) * 2019-10-15 2020-02-21 淮阴工学院 Power grid fault diagnosis method based on multi-feature fusion parameters of wolf pack algorithm
US20210270244A1 (en) * 2020-03-02 2021-09-02 Wuhan University Method and system for fault diagnosis of gearbox of wind turbine generator
CN111665050A (en) * 2020-06-04 2020-09-15 燕山大学 Rolling bearing fault diagnosis method based on clustering K-SVD algorithm
CN113203565A (en) * 2021-03-25 2021-08-03 长江大学 Bearing fault identification method and system based on EEMD sparse decomposition
CN113310684A (en) * 2021-04-20 2021-08-27 东南大学 Gearbox fault feature extraction method based on scale space and improved sparse representation

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
DAI JUNFENG;: "The Design of Digital and Multi-Function Watt-hour Meter", APPLIED MECHANICS AND MATERIALS, no. 392, 23 May 2014 (2014-05-23), pages 750 *
胡泽;张智博;王晓杰;吴雨宸;谢心蕊;: "基于希尔伯特-黄变换和神经网络的滚动轴承故障诊断", 电动工具, no. 01, 18 February 2020 (2020-02-18) *
董健: "风电机组关键部件故障预警与寿命评估方法及其应用研究", 中国博士学位论文全文数据库 (工程科技Ⅱ辑), 15 January 2022 (2022-01-15), pages 042 - 44 *
陈保家: "采用品质因子优化和子带重构的共振稀疏分解滚动轴承故障诊断方法", 西安交通大学学报, vol. 52, no. 04, 14 January 2018 (2018-01-14), pages 70 - 76 *

Also Published As

Publication number Publication date
CN115015752B (en) 2024-08-20

Similar Documents

Publication Publication Date Title
CN109582003B (en) Bearing fault diagnosis method based on pseudo label semi-supervised kernel local Fisher discriminant analysis
Omran et al. Differential evolution methods for unsupervised image classification
CN112084237A (en) Power system abnormity prediction method based on machine learning and big data analysis
CN111709299B (en) Underwater sound target identification method based on weighting support vector machine
CN113177357B (en) Transient stability assessment method for power system
Chen et al. Machine fault classification using deep belief network
CN112257530A (en) Rolling bearing fault diagnosis method based on blind signal separation and support vector machine
CN108647707B (en) Probabilistic neural network creation method, failure diagnosis method and apparatus, and storage medium
CN111681132B (en) Typical power consumption mode extraction method suitable for massive class unbalanced load data
CN113240201B (en) Method for predicting ship host power based on GMM-DNN hybrid model
CN111695611A (en) Bee colony optimization kernel extreme learning and sparse representation mechanical fault identification method
CN114118138A (en) Bearing composite fault diagnosis method based on multi-label field self-adaptive model
CN114091504A (en) Rotary machine small sample fault diagnosis method based on generation countermeasure network
CN116720095A (en) Electrical characteristic signal clustering method for optimizing fuzzy C-means based on genetic algorithm
CN117633688A (en) Large-scale power data anomaly detection method based on ridge regression-k-means clustering-LOF-LSTM fusion algorithm
CN114781551B (en) Battery multi-fault intelligent classification and identification method based on big data
Sharma et al. A semi-supervised generalized vae framework for abnormality detection using one-class classification
CN112362292B (en) Method for anomaly detection of wind tunnel test data
CN114416423A (en) Root cause positioning method and system based on machine learning
CN113705695A (en) Power distribution network fault data identification method based on convolutional neural network
CN115015752A (en) Motor fault diagnosis method based on sparse decomposition and neighborhood bee colony algorithm
CN108898157B (en) Classification method for radar chart representation of numerical data based on convolutional neural network
CN113496255B (en) Power distribution network mixed observation point distribution method based on deep learning and decision tree driving
CN113159132A (en) Hypertension grading method based on multi-model fusion
CN114528906A (en) Fault diagnosis method, device, equipment and medium for rotary machine

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