CN115560966B - Weak fault diagnosis method and system for key components of reinforced sparse filtering fusion motor - Google Patents
Weak fault diagnosis method and system for key components of reinforced sparse filtering fusion motor Download PDFInfo
- Publication number
- CN115560966B CN115560966B CN202211217731.2A CN202211217731A CN115560966B CN 115560966 B CN115560966 B CN 115560966B CN 202211217731 A CN202211217731 A CN 202211217731A CN 115560966 B CN115560966 B CN 115560966B
- Authority
- CN
- China
- Prior art keywords
- fault
- matrix
- multidimensional
- information fusion
- motor
- 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.)
- Active
Links
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
-
- 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
Abstract
The application relates to the field of motor operation and maintenance, and discloses a weak fault diagnosis method and a weak fault diagnosis system for a key component of an enhanced sparse filtering fusion motor, wherein the method comprises the following steps: collecting vibration signals of key parts of a motor, enhancing sparse filtering through different input and output dimensions and utilizingThe norm screens out the corresponding multidimensional fault characteristic information fusion source, and the optimal multidimensional fault characteristic information fusion source is screened out through permutation entropy; extracting and optimizing an inherent manifold by using an improved local tangent space arrangement algorithm, carrying out weighted representation, and carrying out envelope analysis on a representation result to realize fault diagnosis on key components of the motor; the system comprises a data acquisition module, a multidimensional fault feature information fusion source construction module, an optimal multidimensional fault feature information fusion source screening module, an inherent manifold extraction module and a fault diagnosis module. According to the application, the fault impact information of the motor key component is extracted, so that the robustness of the fault feature extraction effect is improved, and the weak fault diagnosis of the motor key component is realized.
Description
Technical Field
The application relates to the technical field of motor operation and maintenance, in particular to a weak fault diagnosis method and system for a reinforced sparse filtering fusion motor key component.
Background
Shafts, bearings, gears are the key components of the motor, and their health status has a critical effect on whether the motor system can function properly. However, because the running environments of the motor are uneven, a large amount of background noise exists in some severe environments, and therefore, fault characteristics generated by key components of the motor cannot be clearly extracted for fault diagnosis.
The traditional motor key component fault characteristic diagnosis method comprises a sound distinguishing method and the like, and the diagnosis method requires a detector to have rich working experience, so that the method has a great disadvantage for workers without working experience. The method still has certain limitations, such as fast inherent component filtering, which can well reflect the population sparsity and service life consistency of the fault features and accords with the fault feature distribution, but has the defect of poor fault feature extraction robustness caused by empirically setting the input dimension and the output dimension of the method; meanwhile, the optimized weight matrix is strictly screened for specific components by using priori knowledge, so that the defect of failure characteristic information loss is also caused.
Disclosure of Invention
Therefore, the technical problem to be solved by the application is to overcome the defects in the prior art, and provide the weak fault diagnosis method and system for enhancing the sparse filtering fusion motor key parts, which can improve the robustness of the fault feature extraction effect, effectively extract abundant fault impact information of the motor key parts and realize weak fault diagnosis of the motor key parts.
In order to solve the technical problems, the application provides a weak fault diagnosis method for a key component of an enhanced sparse filtering fusion motor, which comprises the following steps:
step 1: collecting vibration signals of a motor key component, constructing an initial feature matrix of the motor key component according to the vibration signals, extracting an optimal weight matrix by combining the initial feature matrix with enhanced sparse filtering of different input dimensions and different output dimensions, deconvoluting the optimal weight matrix by using the vibration signals to obtain a fault feature matrix, and utilizing the initial feature matrix to obtain a fault feature matrixThe norm filters the fault feature matrix to obtain multidimensional fault feature information fusion sources corresponding to different input dimensions and output dimensions;
step 2: selecting an optimal multidimensional fault feature information fusion source corresponding to the optimal input dimension and the optimal multidimensional fault feature information fusion source corresponding to the output dimension from multidimensional fault feature information fusion sources corresponding to different input dimensions and different output dimensions through permutation entropy;
step 3: extracting and optimizing an inherent manifold in the optimal multidimensional fault characteristic information fusion source by using a local tangent space arrangement algorithm;
step 4: and carrying out weighted representation on the inherent manifold, and carrying out envelope analysis on the weighted representation result to realize fault diagnosis on key components of the motor.
Preferably, an initial feature matrix of a key component of the motor is constructed according to the vibration signal, specifically:
collecting vibration signals of key parts of motor The vibration signal x is represented as Euclidean space of 1 XN dimension, N represents the number of data points collected;
partitioning x-error a bits into N in N of dimension s Individual segment building momentsArrayWherein N is s ＝N-N in +a，N in Representing an input dimension of enhanced sparse filtering;
randomly generating an initial weight matrixConstructing an initial feature matrix F=WX, < > -corresponding to key components of the motor>Wherein N is out Representing the output dimension of the enhanced sparse filtering.
Preferably, the initial feature matrix and the enhanced sparse filtering of different input dimensions and output dimensions are combined to extract an optimal weight matrix, which specifically comprises:
constructing an objective function C for extracting an optimal weight matrix:
wherein f i j The ith row and jth column elements of the initial feature matrix F are represented, lambda is a weight coefficient, ||w i || 2 =1 represents normalizing each row of the weight matrix W;
minimizing an objective function C by using an L-BFGS algorithm, and taking weight matrixes corresponding to different input dimensions and output dimensions when the objective function C converges as an optimal weight matrix W * 。
Preferably, when the L-BFGS algorithm is used for minimizing the objective function C, the optimized gradient function DeltaW is:
wherein c i Representing the row vector of matrix C, C j The column vectors representing the matrix C are represented,is a matrix with all 1 elements.
Preferably, the use ofThe norm filters the fault feature matrix to obtain multidimensional fault feature information fusion sources corresponding to different input dimensions and output dimensions, and the multidimensional fault feature information fusion sources specifically comprise:
calculating fault feature matrixes F corresponding to different input dimensions and different output dimensions o Each row ofNorm l i ：
Wherein, the liquid crystal display device comprises a liquid crystal display device,representing a fault feature matrix F o I-th, j-th column element of (a);
select F o In (a)Multiple lines with smaller norms form multidimensional fault characteristic information fusion sources H corresponding to different input dimensions and output dimensions o 。
Preferably, the selecting, by permutation entropy, the optimal multidimensional fault feature information fusion source corresponding to the optimal input dimension and the optimal multidimensional fault feature information fusion source corresponding to the output dimension from multidimensional fault feature information fusion sources corresponding to the different input dimension and the different output dimension specifically includes:
calculating multidimensional fault characteristic information fusion sources H corresponding to different input dimensions and different output dimensions o The arrangement entropy of the vector formed by the column mean S of each column
Wherein P is i The probability of the ith permutation of the S sequence reconstruction, and m is the embedding dimension when calculating the permutation entropy value;
taking a multidimensional fault characteristic information fusion source corresponding to the input dimension and the output dimension of the minimum permutation entropy as an optimal multidimensional fault characteristic information fusion source corresponding to the optimal input dimension and the output dimension
Preferably, the local tangent space arrangement algorithm is used for extracting and optimizing the inherent manifold inside the optimal multidimensional fault characteristic information fusion source, and specifically comprises the following steps:
step 3-1: extracting optimal multidimensional fault characteristic information fusion sourceIs the local information V of (2) i ；
Step 3-2: according to the local information V i Calculating a correlation matrix W i According to the correlation matrix W i Constructing an arrangement matrix B;
step 3-3: obtaining fault information U by aligning the full local coordinates of the array matrix B 0 ；
Step 3-4: for the fault information U 0 The abnormal amplitude in the model is subjected to averaging treatment to obtain an inherent manifold
Preferably, fault information U is obtained by aligning the global coordinates of the arrangement matrix B 0 The method specifically comprises the following steps:
step 3-3-1: the first d+1 minimum eigenvalues (lambda) of the permutation matrix B are calculated 1 ,λ 2 ,…,λ d+1 ) Corresponding characteristic directionQuantity (u) 1 ,u 2 ,…,u d+1 )；
Step 3-3-2: selecting feature vectors corresponding to the 2 nd to the (d+1) th minimum feature values to form d-dimensional global coordinates U 0 ＝[u 2 ,u 3 ,…,u d+1 ] T U is set up 0 As fault information.
Preferably, for the fault information U 0 The abnormal amplitude in the model is subjected to averaging treatment to obtain an inherent manifoldThe method comprises the following steps:
inherent manifold shapeThe j-th data point of the i-th-dimensional intrinsic manifold>The calculation method of (1) is as follows:
wherein u is i,j Representing U 0 The j-th data point of the i-th dimension inherent manifold in (a), u i Is U (U) 0 The inherent manifold of the i-th dimension of the (b),is u i N is U 0 Is a total dimension of (2); sigma (sigma) i Is u i S is an outlier determination coefficient.
The application also provides a weak fault diagnosis system for the key components of the enhanced sparse filtering fusion motor, which comprises a data acquisition module, a multidimensional fault characteristic information fusion source construction module, an optimal multidimensional fault characteristic information fusion source screening module, an inherent manifold extraction module and a fault diagnosis module,
the data acquisition module acquires vibration signals of key components of the motor and transmits the vibration signals to the multidimensional fault characteristic information fusion source construction module;
the multidimensional fault feature information fusion source construction module constructs an initial feature matrix of a motor key component according to the vibration signal, combines the initial feature matrix with enhanced sparse filtering of different input dimensions and different output dimensions to extract an optimal weight matrix, deconvolves the optimal weight matrix by using the vibration signal to obtain a fault feature matrix, and utilizes the vibration signal to obtain a fault feature matrixThe norm filters the fault feature matrix to obtain multidimensional fault feature information fusion sources corresponding to different input dimensions and different output dimensions, and the multidimensional fault feature information fusion sources corresponding to the different input dimensions and the different output dimensions are transmitted to the optimal multidimensional fault feature information fusion source filtering module;
the optimal multidimensional fault feature information fusion source screening module screens optimal multidimensional fault feature information fusion sources corresponding to the optimal input dimension and the output dimension from multidimensional fault feature information fusion sources corresponding to different input dimensions and different output dimensions through permutation entropy, and transmits the optimal multidimensional fault feature information fusion sources to the inherent manifold extraction module;
the inherent manifold extracting module uses an improved local tangent space arrangement algorithm to extract and optimize the inherent manifold in the optimal multidimensional fault characteristic information fusion source, and transmits the inherent manifold to the fault diagnosis module;
and the fault diagnosis module performs weighted representation on the inherent manifold, and performs envelope analysis on the weighted representation result to realize fault diagnosis on key components of the motor.
Compared with the prior art, the technical scheme of the application has the following advantages:
(1) The application selects the input dimension of sparse filtering through using permutation entropy self-adaption,Output dimension and utilizeThe norm is used for selecting fault characteristics to form a multidimensional fault characteristic information fusion source, the defect that the robustness of the fault characteristic extraction effect is poor due to the fact that the sparse filtering dimension parameters are selected empirically is overcome, and the robustness of the fault characteristic extraction effect is improved;
(2) The method has the advantages that the fusion source is subjected to nonlinear fusion by improving a local tangent space arrangement algorithm, and the weighted representation is performed after the inherent manifold is optimized, so that fault information is richer, the influence of abnormal amplitude of fusion results on frequency domain representation and the one-sided property of single fusion components are overcome, the fault impact information of the motor key components can be effectively extracted, the fault characteristics of the motor key components are more clearly and accurately extracted, and weak fault diagnosis of the motor key components is realized.
Drawings
In order that the application may be more readily understood, a more particular description of the application will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings, in which:
figure 1 is a flow chart of the present application,
figure 2 is a graph of bearing outer race fault vibration data collected in an embodiment of the present application,
FIG. 3 is a schematic diagram of selecting optimal input dimensions and output dimensions according to minimum permutation entropy in an embodiment of the present application,
figure 4 is a schematic diagram of selecting neighbor points according to minimum permutation entropy in an embodiment of the present application,
fig. 5 is a graph of weighted representation results and envelope analysis obtained in an embodiment of the present application.
Detailed Description
The present application will be further described with reference to the accompanying drawings and specific examples, which are not intended to be limiting, so that those skilled in the art will better understand the application and practice it.
In the description of the present application, it is to be understood that the meaning of "multiple" is two or more, unless explicitly defined otherwise. Furthermore, the term "comprising" is intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements but may, optionally, include other steps or elements not listed or inherent to such process, method, article, or apparatus.
Referring to a flow chart of fig. 1, the application discloses a weak fault diagnosis method for a key component of an enhanced sparse filtering fusion motor, which comprises the following steps:
step 1: collecting vibration signals of a motor key component, constructing an initial feature matrix of the motor key component according to the vibration signals, extracting an optimal weight matrix by combining the initial feature matrix with enhanced sparse filtering of different input dimensions and different output dimensions, deconvoluting the optimal weight matrix by using the vibration signals to obtain a fault feature matrix, and utilizing the initial feature matrix to obtain a fault feature matrixAnd the norm filters the fault feature matrix to obtain multidimensional fault feature information fusion sources corresponding to different input dimensions and output dimensions, so that the problems of information redundancy, interference information and the like of a fault signal enhancement matrix with high output dimensions when the input dimensions and the output dimensions are overlarge are avoided.
Step 1-1: collecting vibration signals of key parts of the motor, and constructing an initial feature matrix of the key parts of the motor according to the vibration signals:
step 1-1-1: collecting vibration signals of key parts of motor The vibration signal x is represented as Euclidean space of 1 XN dimension, N represents the number of data points collected;
step 1-1-2: partitioning x-error a bits into N in N of dimension s Individual segment constitutionMatrix arrayWherein N is s ＝N-N in +a，N in Representing an input dimension of enhanced sparse filtering; in this embodiment a=1.
Step 1-1-3: randomly generating an initial weight matrixConstructing an initial feature matrix F=WX, < > -corresponding to key components of the motor>Wherein N is out Representing the output dimension of the enhanced sparse filtering.
For the selection of input dimension and output dimension, setting the input dimension N in Equal to the output dimension N out Simultaneous optimization of N can be avoided in And N out The two parameters bring large calculation burden and overcome the problem of information loss caused by failure characteristic output dimension reduction, and the information loss is verified through experiments, when N in The fast inherent filtering can obtain the fault characteristic enhancement effect at the time of more than 20, but the effect is uneven. Thus, N is set in the present embodiment out ＝N in ，N in ∈[20,200]Taking 10 as a step length to take a value.
Step 1-2: extracting an optimal weight matrix by combining the initial feature matrix with enhanced sparse filtering of different input dimensions and output dimensions:
step 1-2-1: constructing an objective function C for extracting an optimal weight matrix:
wherein f i j The ith row and jth column elements of the initial feature matrix F are represented, lambda is a weight coefficient, ||w i || 2 =1 represents normalizing each row of the weight matrix W; through multiple tests, the fault characteristics have good extraction effect when the weight coefficient is more than 0.01 and less than or equal to 1Thus λ=1 in this embodiment.
Step 1-2-2: the L-BFGS algorithm is used for minimizing the objective function C, and the optimized gradient function DeltaW when the L-BFGS algorithm is used for minimizing the objective function C is as follows:
wherein c i Representing the row vector of matrix C, C j The column vectors representing the matrix C are represented,a matrix with all 1 elements;
step 1-2-3: taking weight matrixes corresponding to different input dimensions and output dimensions when the objective function C converges as an optimal weight matrix W * 。
Step 1-3: deconvolution is carried out on the optimal weight matrix by using the vibration signal to obtain a fault feature matrix: an optimal weight matrix W corresponding to different input dimensions and output dimensions is obtained * Convolving each row of the motor key component vibration signals x acquired to obtain fault feature matrixes F corresponding to different input dimensions and output dimensions o 。
Step 1-4: by means ofAnd screening the fault feature matrix by using norms to obtain multidimensional fault feature information fusion sources corresponding to different input dimensions and output dimensions:
step 1-4-1: calculating fault feature matrixes F corresponding to different input dimensions and different output dimensions o Each row ofNorm l i ：
In the process, the liquid crystal display device comprises a liquid crystal display device,representing a fault feature matrix F o The ith row, the jth column element;
step 1-4-2: select F o In (a)N rows with smaller norms form multidimensional fault characteristic information fusion sources H corresponding to different input dimensions and output dimensions o . Multiple experiments find that selection F o Middle->10-20 lines with smaller norms form multidimensional fault characteristic information fusion sources H corresponding to different input dimensions and output dimensions o Further, the subsequent steps are performed, which can finally obtain a good fault feature extraction effect, so n=15 in this embodiment.
Step 2: the optimal input dimension and the optimal output dimension are adaptively screened from the multidimensional fault feature information fusion source through permutation entropy, so that the problem of poor robustness of a fault feature extraction effect caused by empirically selecting dimension parameters is avoided; and obtaining a corresponding optimal multidimensional fault characteristic information fusion source according to the optimal input dimension and the optimal output dimension.
Step 2-1: calculating multidimensional fault characteristic information fusion sources H corresponding to different input dimensions and different output dimensions o The arrangement entropy of the vector formed by the column mean S of each column
Wherein P is i Is the probability of the ith permutation of S sequence reconstruction, s=mean (H o 1), mean (, 1) represents the averaging function; m is the embedding dimension when calculating the permutation entropy;
step 2-2: taking a multidimensional fault characteristic information fusion source corresponding to the input dimension and the output dimension of the minimum permutation entropy as an optimal multidimensional fault characteristic information fusion source corresponding to the optimal input dimension and the output dimension
Step 3: the improved local tangent space arrangement algorithm is used for extracting and optimizing the inherent manifold in the optimal multidimensional fault characteristic information fusion source, so that the problem of fault characteristic information loss caused by screening a weight matrix through priori knowledge is avoided; the optimization of the inherent manifold is specifically to perform standardized processing on the abnormal amplitude of the inherent manifold, so that the influence of the abnormal amplitude of the fusion result on the frequency domain representation is overcome.
Step 3-1: extracting optimal multidimensional fault characteristic information fusion sourceIs the local information V of (2) i ：
Step 3-1-1: fusion source for optimal multidimensional fault characteristic informationEach column h of (3) i Determining a permutation entropy comprising h i The k neighboring points in the inner form a data set Z i ；
Step 3-1-2: centralized dataset Z i Obtaining a matrixWherein (1)>Is Z i Mean value of e k Column vectors for all elements of 1;
step 3-1-3: calculation ofRight singular vector formation corresponding to the first d largest singular valuesMatrix V i ；
Step 3-2: according to the local information V i Calculating a correlation matrix W i According to the correlation matrix W i Constructing an arrangement matrix B:
step 3-2-1: according toAnd Z i Calculating to obtain 0-1 selection matrix S i The method comprises the following steps: />
Step 3-2-2: calculate V i Is related to the matrix W of (2) i The method comprises the following steps:
step 3-2-3: binding S i And W is i The configuration array matrix B is:
step 3-3: obtaining fault information U by aligning the full local coordinates of the array matrix B 0 ：
Step 3-3-1: the first d+1 minimum eigenvalues (lambda) of the permutation matrix B are calculated 1 ,λ 2 ,…,λ d+1 ) And its corresponding feature vector (u) 1 ,u 2 ,…,u d+1 )；
Step 3-3-2: selecting feature vectors corresponding to the 2 nd to the (d+1) th minimum feature values to form d-dimensional global coordinates U 0 ＝[u 2 ,u 3 ,…,u d+1 ] T ，U 0 Is thatDimension reduced form, U 0 As fault information; an excessively large setting of the inherent dimension d generates a lot of redundant information, and an excessively small setting of useful information is ignored in performing dimension reduction, so d=3 is taken in this embodiment.
Step 3-4: for the reasons described aboveAnd carrying out averaging treatment on the abnormal amplitude values in the barrier information to obtain an inherent manifold. In particular using fault information U 0 Calibration of fault information U by standard deviation of intrinsic manifolds of each dimension 0 The abnormal amplitude in (a) is obtained to obtain an inherent manifold
Inherent manifold shapeThe j-th data point of the i-th-dimensional intrinsic manifold>The calculation method of (1) is as follows:
wherein u is i,j Representing U 0 The j-th data point of the i-th dimension inherent manifold in (a), u i Is U (U) 0 The inherent manifold of the i-th dimension of the (b),is u i N is U 0 Is a total dimension of (2); sigma (sigma) i Is u i S is an outlier determination coefficient, and in this embodiment, s is uniformly valued at 6.
Step 4: the inherent manifold is weighted and represented, so that the one-sidedness of a single fusion component is avoided, and the final fault representation contains richer fault characteristic information; and carrying out envelope analysis on the weighted representation result to realize fault diagnosis on key components of the motor.
Step 4-1: and carrying out weighted representation on the inherent manifold, specifically:
for inherent manifoldIntrinsic manifold of the ith dimension +.>The weighted representation is performed to obtain a result vector R as:
wherein lambda is i Is u i The corresponding characteristic value.
Step 4-2: and carrying out envelope analysis on the weighted representation result to realize fault diagnosis on key components of the motor.
The application also discloses a weak fault diagnosis system for the key components of the reinforced sparse filtering fusion motor, which comprises a data acquisition module, a multidimensional fault characteristic information fusion source construction module, an optimal multidimensional fault characteristic information fusion source screening module, an inherent manifold extraction module and a fault diagnosis module.
The data acquisition module acquires vibration signals of key components of the motor and transmits the vibration signals to the multidimensional fault characteristic information fusion source construction module; the multidimensional fault feature information fusion source construction module constructs an initial feature matrix of a motor key component according to the vibration signal, combines the initial feature matrix with enhanced sparse filtering with different input dimensions and output dimensions to extract an optimal weight matrix, deconvolves the optimal weight matrix by using the vibration signal to obtain a fault feature matrix, and utilizes the vibration signal to obtain a fault feature matrixThe norm filters the fault feature matrix to obtain multidimensional fault feature information fusion sources corresponding to different input dimensions and different output dimensions, and the multidimensional fault feature information fusion sources corresponding to the different input dimensions and the different output dimensions are transmitted to the optimal multidimensional fault feature information fusion source filtering module; the optimal multidimensional fault characteristic information fusion source screening module is used for selecting the optimal multidimensional fault characteristic information fusion source through permutation entropyScreening out optimal multidimensional fault feature information fusion sources corresponding to the optimal input dimension and the optimal multidimensional fault feature information fusion sources corresponding to the output dimension from multidimensional fault feature information fusion sources corresponding to different input dimensions and different output dimensions, and transmitting the optimal multidimensional fault feature information fusion sources to the inherent manifold extraction module; the inherent manifold extracting module uses an improved local tangent space arrangement algorithm to extract and optimize the inherent manifold in the optimal multidimensional fault characteristic information fusion source, and transmits the inherent manifold to the fault diagnosis module; and the fault diagnosis module performs weighted representation on the inherent manifold, and performs envelope analysis on the weighted representation result to realize fault diagnosis on key components of the motor.
The application selects the input dimension and the output dimension of sparse filtering by using permutation entropy self-adaption and utilizesThe norm is used for selecting fault characteristics to form a multidimensional fault characteristic information fusion source, the defect that the robustness of the fault characteristic extraction effect is poor due to the fact that the sparse filtering dimension parameters are selected empirically is overcome, and the robustness of the fault characteristic extraction effect is improved. Meanwhile, the fusion source is subjected to nonlinear fusion by improving a local tangent space arrangement algorithm, and the fusion source is subjected to weighted representation after the inherent manifold is optimized, so that fault information is more abundant, the influence of abnormal amplitude of fusion results on frequency domain representation and one-sided performance of single fusion components are overcome, ideal fault impact information of key components of the motor can be effectively obtained, the fault characteristics of the key components of the motor are more clearly and accurately extracted, and weak fault diagnosis of the key components of the motor is realized.
In order to further illustrate the beneficial effects of the application, in this embodiment, a set of fault vibration data of the outer ring of the bearing as shown in fig. 2 is collected, the upper half part of fig. 2 is an original vibration signal of a key component of the motor, and the lower half part is an envelope spectrum obtained by enveloping the original vibration signal. As shown in fig. 3, the permutation entropy value is minimum when the input dimension and the output dimension are 40, so that the multi-dimensional fault feature information fusion source when the input dimension and the output dimension are 40 is selected as the optimal multi-dimensional fault feature information fusion source. As shown in fig. 4, since the permutation entropy of the optimal multidimensional fault signature fusion source is minimum when the number k of neighboring points is 15, the inherent manifold when k is 15 is expressed by weighting. The upper half of fig. 5 is a diagram showing the result vector R obtained by weighting, and the lower half of fig. 5 is an envelope spectrum obtained by enveloping R. As can be seen from fig. 5, the time domain waveform of the bearing fault impact is clearly discernable, the characteristic frequency of the fault impact and the frequency multiplication relation thereof are also clearly discernable, and the fault diagnosis performed on the basis is more accurate, thereby proving the beneficial effects of the application.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It is apparent that the above examples are given by way of illustration only and are not limiting of the embodiments. Other variations and modifications of the present application will be apparent to those of ordinary skill in the art in light of the foregoing description. It is not necessary here nor is it exhaustive of all embodiments. And obvious variations or modifications thereof are contemplated as falling within the scope of the present application.
Claims (8)
1. A weak fault diagnosis method for a key component of an enhanced sparse filtering fusion motor is characterized by comprising the following steps:
step 1: collecting vibration signals of a motor key component, wherein the motor key component comprises a shaft, a bearing and a gear, an initial feature matrix of the motor key component is constructed according to the vibration signals, an optimal weight matrix is extracted by combining the initial feature matrix with enhanced sparse filtering of different input dimensions and different output dimensions, the optimal weight matrix is deconvoluted by using the vibration signals to obtain a fault feature matrix, and the fault feature matrix is obtained by using the vibration signalsThe norm filters the fault feature matrix to obtain multidimensional fault feature information fusion sources corresponding to different input dimensions and output dimensions;
the initial feature matrix of the key parts of the motor is constructed according to the vibration signals, and the initial feature matrix is specifically as follows: collecting vibration signals of key parts of motor The vibration signal x is represented as Euclidean space of 1 XN dimension, N represents the number of data points collected; partitioning x-error a bits into N in N of dimension s The individual fragments form a matrix->Wherein N is s ＝N-N in +a，N in Representing an input dimension of enhanced sparse filtering; randomly generating an initial weight matrix->Constructing an initial feature matrix F=WX, < > -corresponding to key components of the motor>Wherein N is out Representing the output dimension of the enhanced sparse filtering;
the utilization ofThe norm filters the fault feature matrix to obtain multidimensional fault feature information fusion sources corresponding to different input dimensions and output dimensions, and the multidimensional fault feature information fusion sources specifically comprise: calculating fault feature matrixes F corresponding to different input dimensions and different output dimensions o +.>Norm l i ：/>Wherein (1)>Representing characteristic moment of failureArray F o I-th, j-th column element of (a); select F o Middle->Multiple lines with smaller norms form multidimensional fault characteristic information fusion sources H corresponding to different input dimensions and output dimensions o ；
Step 2: selecting an optimal multidimensional fault feature information fusion source corresponding to the optimal input dimension and the optimal multidimensional fault feature information fusion source corresponding to the output dimension from multidimensional fault feature information fusion sources corresponding to different input dimensions and different output dimensions through permutation entropy;
step 3: extracting and optimizing an inherent manifold in the optimal multidimensional fault characteristic information fusion source by using a local tangent space arrangement algorithm;
step 4: and carrying out weighted representation on the inherent manifold, and carrying out envelope analysis on the weighted representation result to realize fault diagnosis on key components of the motor.
2. The weak fault diagnosis method for the key components of the enhanced sparse filtering fusion motor according to claim 1 is characterized by comprising the following steps: combining the initial feature matrix with the enhanced sparse filtering of different input dimensions and output dimensions to extract an optimal weight matrix, wherein the optimal weight matrix comprises the following specific steps of:
constructing an objective function C for extracting an optimal weight matrix:
wherein f i j The ith row and jth column elements of the initial feature matrix F are represented, lambda is a weight coefficient, ||w i || 2 =1 represents normalizing each row of the weight matrix W;
minimizing an objective function C by using an L-BFGS algorithm, and taking weight matrixes corresponding to different input dimensions and output dimensions when the objective function C converges as an optimal weight matrix W * 。
3. The weak fault diagnosis method for the key components of the enhanced sparse filtering fusion motor according to claim 2, which is characterized by comprising the following steps: when the L-BFGS algorithm is used for minimizing the objective function C, the optimized gradient function delta W is as follows:
wherein c i Representing the row vector of matrix C, C j The column vectors representing the matrix C are represented,is a matrix with all 1 elements.
4. The weak fault diagnosis method for the key components of the enhanced sparse filtering fusion motor according to claim 1 is characterized by comprising the following steps: the optimal multidimensional fault feature information fusion source corresponding to the optimal input dimension and the optimal multidimensional fault feature information fusion source corresponding to the output dimension are screened out from the multidimensional fault feature information fusion sources corresponding to the different input dimension and the different output dimension through permutation entropy, and specifically comprises the following steps:
calculating multidimensional fault characteristic information fusion sources H corresponding to different input dimensions and different output dimensions o The arrangement entropy of the vector formed by the column mean S of each column
Wherein P is i The probability of the ith permutation of the S sequence reconstruction, and m is the embedding dimension when calculating the permutation entropy value;
taking a multidimensional fault characteristic information fusion source corresponding to the input dimension and the output dimension of the minimum permutation entropy as an optimal multidimensional fault characteristic information fusion source corresponding to the optimal input dimension and the output dimension
5. The weak fault diagnosis method for the key components of the reinforced sparse filtering fusion motor according to any one of claims 1-4, which is characterized by comprising the following steps: the method comprises the steps of extracting and optimizing the inherent manifold in the optimal multidimensional fault characteristic information fusion source by using a local tangent space arrangement algorithm, and specifically comprises the following steps:
step 3-1: extracting optimal multidimensional fault characteristic information fusion sourceIs the local information V of (2) i ；
Step 3-2: according to the local information V i Calculating a correlation matrix W i According to the correlation matrix W i Constructing an arrangement matrix B;
step 3-3: obtaining fault information U by aligning the full local coordinates of the array matrix B 0 ；
Step 3-4: for the fault information U 0 The abnormal amplitude in the model is subjected to averaging treatment to obtain an inherent manifold
6. The weak fault diagnosis method for the key components of the enhanced sparse filtering fusion motor, which is disclosed by claim 5, is characterized in that: obtaining fault information U by aligning the full local coordinates of the array matrix B 0 The method specifically comprises the following steps:
step 3-3-1: the first d+1 minimum eigenvalues (lambda) of the permutation matrix B are calculated 1 ,λ 2 ,…,λ d+1 ) Corresponding feature vector (u) 1 ,u 2 ,…,u d+1 )；
Step 3-3-2: selecting feature vectors corresponding to the 2 nd to the (d+1) th minimum feature values to form d-dimensional global coordinates U 0 ＝[u 2 ,u 3 ,…,u d+1 ] T U is set up 0 As fault information。
7. The weak fault diagnosis method for the key components of the enhanced sparse filtering fusion motor, which is disclosed by claim 5, is characterized in that: for the fault information U 0 The abnormal amplitude in the model is subjected to averaging treatment to obtain an inherent manifoldThe method comprises the following steps:
inherent manifold shapeThe j-th data point of the i-th-dimensional intrinsic manifold>The calculation method of (1) is as follows:
wherein u is i,j Representing U 0 The j-th data point of the i-th dimension inherent manifold in (a), u i Is U (U) 0 The inherent manifold of the i-th dimension of the (b),is u i N is U 0 Is a total dimension of (2); sigma (sigma) i Is u i S is an outlier determination coefficient.
8. A weak fault diagnosis system for a key component of an enhanced sparse filtering fusion motor is characterized in that: comprises a data acquisition module, a multidimensional fault characteristic information fusion source construction module, an optimal multidimensional fault characteristic information fusion source screening module, an inherent manifold extraction module and a fault diagnosis module,
the data acquisition module acquires vibration signals of key parts of the motor, wherein the key parts of the motor comprise shafts, bearings and gears, and the vibration signals are transmitted to the multidimensional fault characteristic information fusion source construction module;
the multidimensional fault feature information fusion source construction module constructs an initial feature matrix of a motor key component according to the vibration signal, combines the initial feature matrix with enhanced sparse filtering of different input dimensions and different output dimensions to extract an optimal weight matrix, deconvolves the optimal weight matrix by using the vibration signal to obtain a fault feature matrix, and utilizes the vibration signal to obtain a fault feature matrixThe norm filters the fault feature matrix to obtain multidimensional fault feature information fusion sources corresponding to different input dimensions and different output dimensions, and the multidimensional fault feature information fusion sources corresponding to the different input dimensions and the different output dimensions are transmitted to the optimal multidimensional fault feature information fusion source filtering module;
the initial feature matrix of the key parts of the motor is constructed according to the vibration signals, and the initial feature matrix is specifically as follows: collecting vibration signals of key parts of motor The vibration signal x is represented as Euclidean space of 1 XN dimension, N represents the number of data points collected; partitioning x-error a bits into N in N of dimension s The individual fragments form a matrix->Wherein N is s ＝N-N in +a，N in Representing an input dimension of enhanced sparse filtering; randomly generating an initial weight matrix->Constructing an initial feature matrix F=WX, < > -corresponding to key components of the motor>Wherein N is out Representing the output dimension of the enhanced sparse filtering;
the utilization ofThe norm filters the fault feature matrix to obtain multidimensional fault feature information fusion sources corresponding to different input dimensions and output dimensions, and the multidimensional fault feature information fusion sources specifically comprise: calculating fault feature matrixes F corresponding to different input dimensions and different output dimensions o +.>Norm l i ：/>Wherein (1)>Representing a fault feature matrix F o I-th, j-th column element of (a); select F o Middle->Multiple lines with smaller norms form multidimensional fault characteristic information fusion sources H corresponding to different input dimensions and output dimensions o ；
The optimal multidimensional fault feature information fusion source screening module screens optimal multidimensional fault feature information fusion sources corresponding to the optimal input dimension and the output dimension from multidimensional fault feature information fusion sources corresponding to different input dimensions and different output dimensions through permutation entropy, and transmits the optimal multidimensional fault feature information fusion sources to the inherent manifold extraction module;
the inherent manifold extracting module uses an improved local tangent space arrangement algorithm to extract and optimize the inherent manifold in the optimal multidimensional fault characteristic information fusion source, and transmits the inherent manifold to the fault diagnosis module;
and the fault diagnosis module performs weighted representation on the inherent manifold, and performs envelope analysis on the weighted representation result to realize fault diagnosis on key components of the motor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211217731.2A CN115560966B (en) | 2022-09-30 | 2022-09-30 | Weak fault diagnosis method and system for key components of reinforced sparse filtering fusion motor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211217731.2A CN115560966B (en) | 2022-09-30 | 2022-09-30 | Weak fault diagnosis method and system for key components of reinforced sparse filtering fusion motor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115560966A CN115560966A (en) | 2023-01-03 |
| CN115560966B true CN115560966B (en) | 2023-09-15 |
Family
ID=84744690
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211217731.2A Active CN115560966B (en) | 2022-09-30 | 2022-09-30 | Weak fault diagnosis method and system for key components of reinforced sparse filtering fusion motor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115560966B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103533349A (en) * | 2013-09-26 | 2014-01-22 | 广东电网公司电力科学研究院 | Support vector machine-based fast inter-frame prediction macro block mode selection method for B frame |
| CN108181107A (en) * | 2018-01-12 | 2018-06-19 | 东北电力大学 | The Wind turbines bearing mechanical method for diagnosing faults of meter and more class objects |
| CN109655259A (en) * | 2018-11-23 | 2019-04-19 | 华南理工大学 | Combined failure diagnostic method and device based on depth decoupling convolutional neural networks |
| CN109656662A (en) * | 2018-12-14 | 2019-04-19 | 浙江财经大学 | The visualization aligning method of parallel coordinate axes based on cluster correlation |
| CN110770561A (en) * | 2017-06-12 | 2020-02-07 | 利乐拉瓦尔集团及财务有限公司 | Method for predicting failure of machine part moving periodically |
| CN114282571A (en) * | 2021-10-27 | 2022-04-05 | 河北建投能源投资股份有限公司 | Method, system, equipment and medium for constructing multidimensional health index of bearing |
| CN114444552A (en) * | 2022-04-08 | 2022-05-06 | 新黎明科技股份有限公司 | Weak fault diagnosis method, medium, electronic equipment and system for explosion-proof motor bearing |
| CN114608826A (en) * | 2022-03-16 | 2022-06-10 | 合肥工业大学 | Training method, diagnosis method and diagnosis device of bearing fault diagnosis model |
| CN114742115A (en) * | 2022-06-13 | 2022-07-12 | 西南交通大学 | Rolling bearing fault diagnosis model and method based on temperature and vibration characteristic fusion |
| CN115049267A (en) * | 2022-06-20 | 2022-09-13 | 华电电力科学研究院有限公司 | Wind turbine generator data processing method, system, equipment and computer medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7945397B2 (en) * | 2009-04-02 | 2011-05-17 | Honeywell International Inc. | System and method for gearbox health monitoring |
| US8963733B2 (en) * | 2012-02-13 | 2015-02-24 | Honeywell International Inc. | System and method for blind fault detection for rotating machinery |
| US11320552B2 (en) * | 2018-03-29 | 2022-05-03 | Exxonmobil Upstream Research Company | Enhanced surveillance of subsurface operation integrity using neural network analysis of microseismic data |
| CN108972558B (en) * | 2018-08-16 | 2020-02-21 | 居鹤华 | Multi-axis robot dynamics modeling method based on axis invariants |
-
2022
- 2022-09-30 CN CN202211217731.2A patent/CN115560966B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103533349A (en) * | 2013-09-26 | 2014-01-22 | 广东电网公司电力科学研究院 | Support vector machine-based fast inter-frame prediction macro block mode selection method for B frame |
| CN110770561A (en) * | 2017-06-12 | 2020-02-07 | 利乐拉瓦尔集团及财务有限公司 | Method for predicting failure of machine part moving periodically |
| CN108181107A (en) * | 2018-01-12 | 2018-06-19 | 东北电力大学 | The Wind turbines bearing mechanical method for diagnosing faults of meter and more class objects |
| CN109655259A (en) * | 2018-11-23 | 2019-04-19 | 华南理工大学 | Combined failure diagnostic method and device based on depth decoupling convolutional neural networks |
| CN109656662A (en) * | 2018-12-14 | 2019-04-19 | 浙江财经大学 | The visualization aligning method of parallel coordinate axes based on cluster correlation |
| CN114282571A (en) * | 2021-10-27 | 2022-04-05 | 河北建投能源投资股份有限公司 | Method, system, equipment and medium for constructing multidimensional health index of bearing |
| CN114608826A (en) * | 2022-03-16 | 2022-06-10 | 合肥工业大学 | Training method, diagnosis method and diagnosis device of bearing fault diagnosis model |
| CN114444552A (en) * | 2022-04-08 | 2022-05-06 | 新黎明科技股份有限公司 | Weak fault diagnosis method, medium, electronic equipment and system for explosion-proof motor bearing |
| CN114742115A (en) * | 2022-06-13 | 2022-07-12 | 西南交通大学 | Rolling bearing fault diagnosis model and method based on temperature and vibration characteristic fusion |
| CN115049267A (en) * | 2022-06-20 | 2022-09-13 | 华电电力科学研究院有限公司 | Wind turbine generator data processing method, system, equipment and computer medium |
Non-Patent Citations (1)
| Title |
|---|
| 稀疏优化在旋转机械初始故障诊断中的应用;高杨德;《中国优秀硕士学位论文全文数据库 (工程科技Ⅱ辑)》;C029-137 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115560966A (en) | 2023-01-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111723732B (en) | Optical remote sensing image change detection method, storage medium and computing equipment | |
| CN113159051B (en) | Remote sensing image lightweight semantic segmentation method based on edge decoupling | |
| CN110297479B (en) | Hydroelectric generating set fault diagnosis method based on convolutional neural network information fusion | |
| CN114216682B (en) | Service life prediction method and device of rolling bearing based on TCN and BLS | |
| CN110807544B (en) | Oil field residual oil saturation distribution prediction method based on machine learning | |
| CN111351665B (en) | Rolling bearing fault diagnosis method based on EMD and residual error neural network | |
| CN113112446A (en) | Tunnel surrounding rock level intelligent judgment method based on residual convolutional neural network | |
| JP2019204321A (en) | Information processing method, information processor, and program | |
| Fang et al. | Multiscale CNNs ensemble based self-learning for hyperspectral image classification | |
| CN114548190A (en) | Wind turbine fault diagnosis method based on self-adaptive residual error neural network | |
| CN116416497A (en) | Bearing fault diagnosis system and method | |
| CN114565594A (en) | Image anomaly detection method based on soft mask contrast loss | |
| CN115688040A (en) | Mechanical equipment fault diagnosis method, device, equipment and readable storage medium | |
| CN115560966B (en) | Weak fault diagnosis method and system for key components of reinforced sparse filtering fusion motor | |
| CN103617119A (en) | Method for selecting regression test cases based on statement interactive coverage | |
| CN116994044A (en) | Construction method of image anomaly detection model based on mask multi-mode generation countermeasure network | |
| CN112182568B (en) | Malicious code classification based on graph convolution network and topic model | |
| CN113112447A (en) | Tunnel surrounding rock grade intelligent determination method based on VGG convolutional neural network | |
| CN117150959A (en) | Super-resolution reconstruction-based flow field data compression and restoration method | |
| CN116222753A (en) | Rotor system fault sensitivity feature extraction method and system | |
| CN116595465A (en) | High-dimensional sparse data outlier detection method and system based on self-encoder and data enhancement | |
| CN111027669A (en) | Method and device for realizing deep neural network on field programmable gate array | |
| CN113435321A (en) | Method, system and equipment for evaluating state of main shaft bearing and readable storage medium | |
| CN113627538A (en) | Method and electronic device for training asymmetric generation countermeasure network to generate image | |
| CN112329013A (en) | Malicious code classification method based on graph convolution network and topic model |
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 |