A method of bearing fault classification and related apparatus
By constructing a sparse dictionary and calculating sparse coefficients through a dictionary learning algorithm, the nonlinearity and information redundancy problems of bearing fault signals are solved, and fault feature extraction and accurate classification are achieved under low signal-to-noise ratio conditions.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- PETROCHINA CO LTD
- Filing Date
- 2022-08-09
- Publication Date
- 2026-06-23
AI Technical Summary
In complex mechanical equipment operating environments, early bearing fault signals have nonlinear and information redundancy characteristics, making it difficult to effectively extract fault features from low signal-to-noise ratio vibration signals, which leads to difficulties in the classification and diagnosis of rolling bearing faults.
A sparse dictionary is constructed using a dictionary learning algorithm. By solving the sparse optimization problem and calculating the sparse coefficients, the signal is reconstructed and the correlation is compared to achieve accurate identification of fault types.
Accurately extracting fault features from low signal-to-noise ratio vibration signals reduces noise interference and measurement point location limitations, thereby improving the accuracy of early fault identification.
Smart Images
Figure CN117648598B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of fault detection technology, and in particular to a method and related apparatus for classifying bearing faults. Background Technology
[0002] Rolling bearings are key components in rotating machinery systems such as pumps and motors. Due to their long service life and high rotational speed, rolling bearings are highly susceptible to failures such as cracks and pitting. Bearing failures account for approximately 30% of all rotating machinery malfunctions, and the operating condition of the bearings directly affects the performance of the entire machine. Therefore, accurate diagnosis of rolling bearing failures is crucial.
[0003] In actual industrial production, due to factors such as the complex operating environment of mechanical equipment, significant noise interference, and limited measurement point locations, early bearing fault signals exhibit nonlinear and redundant characteristics, making it difficult to obtain fault information. How to effectively extract fault features from low signal-to-noise ratio vibration signals, and thus achieve rolling bearing fault classification and diagnosis, is a critical technical problem that urgently needs to be solved. Summary of the Invention
[0004] In view of the above problems, the present invention is proposed to provide a method and related apparatus for bearing fault classification that overcomes or at least partially solves the above problems.
[0005] In a first aspect, embodiments of the present invention provide a bearing fault classification method, including:
[0006] After processing the first vibration signals of the bearing under multiple known fault types, the collected signals are input into a preset dictionary learning algorithm model to construct a sparse dictionary containing training samples of multiple fault types.
[0007] After processing the second vibration signal to be classified into faults, a test sample is obtained. The test sample is then subjected to sparse optimization to obtain the first sparse coefficient.
[0008] Based on the first sparse coefficients and training samples under multiple fault types, reconstruction is performed to obtain reconstructed signals corresponding to multiple fault types.
[0009] The test sample is compared with the reconstructed signals corresponding to the multiple fault types, and the fault type corresponding to the test sample is determined based on the comparison results.
[0010] In one embodiment, the first vibration signal collected under multiple known fault types of the bearing is processed and then input into a preset dictionary learning algorithm model to construct a sparse dictionary containing training samples of multiple fault types, including:
[0011] The sparse coefficients and dictionary are alternately calculated using a dictionary learning algorithm model until the dictionary and sparse coefficients no longer change. The iteration is then stopped, resulting in the sparse dictionary containing training samples of multiple fault types.
[0012] In one embodiment, the test samples are calculated using sparse optimization to obtain first sparse coefficients, including:
[0013] Calculate the second sparsity coefficient and the RCMFE weighting parameter for the test samples respectively;
[0014] Calculate the weighting coefficients based on the RCMFE weighting parameters;
[0015] Based on the second sparsity coefficient and the weighting coefficient, the test sample is sparsely optimized to obtain the first sparsity coefficient.
[0016] In one embodiment, a weighting coefficient is calculated based on the RCMFE weighting parameter, and sparse optimization is performed on the test samples based on the second sparse coefficient and the weighting coefficient to obtain the first sparse coefficient, including:
[0017] The weighting coefficients are calculated using the following formula:
[0018] W(i)=A*|(hH(i)) n |;
[0019] In the above formula, h and H(i) represent the RCMFE weighting parameters of the test sample and the training sample under different fault types, respectively, A,n is a constant, and i is the fault type number. The weighting coefficients are calculated as follows:
[0020] w(i) = 1 / W(i);
[0021] The first sparsity coefficient is obtained by solving the sparsity optimization problem of the test samples using the following formula:
[0022]
[0023] in, ε is the first sparse coefficient, w is the weighting coefficient, D is the training sample in the sparse dictionary, and ε is the sparse representation residual error.
[0024] In one embodiment, the test sample is compared with the reconstructed signals corresponding to the plurality of fault types, and the fault type corresponding to the test sample is determined based on the comparison result, including:
[0025] Determine the correlation between the test sample and the reconstructed signals corresponding to the multiple fault types, and determine the fault type corresponding to the test sample based on the correlation result.
[0026] In one embodiment, the correlation between the test sample and the reconstructed signals corresponding to the plurality of fault types is determined by the following formula, and the fault type corresponding to the test sample is determined based on the result of the correlation:
[0027] Calculate the correlation using the cross-correlation function formula:
[0028]
[0029] In the above formula, N is the signal dimension, τ is the time shift deviation, and x(t) and y(t) are the signals with time shift deviations, respectively; R xy (t) represents the correlation between the test sample and the reconstructed signal;
[0030] The fault type corresponding to the reconstructed signal with the highest correlation is taken as the fault type of the test sample.
[0031] In one embodiment: processing the first vibration signal and processing the second vibration signal include:
[0032] The first and second vibration signals are segmented and then normalized in amplitude.
[0033] Secondly, embodiments of the present invention provide a bearing fault classification device, comprising:
[0034] The sparse dictionary construction module is used to process the first vibration signal of the bearing under multiple known fault types and input it into the preset dictionary learning algorithm model to construct a sparse dictionary containing training samples of multiple fault types.
[0035] The first sparse coefficient solving module is used to process the second vibration signal to be classified into faults to obtain test samples, and to perform sparse optimization solving on the test samples to obtain the first sparse coefficients.
[0036] The signal reconstruction module is used to reconstruct the signals based on the first sparse coefficients and training samples under multiple fault types to obtain reconstructed signals corresponding to multiple fault types.
[0037] The fault classification module is used to compare the test sample with the reconstructed signals corresponding to the multiple fault types, and determine the fault type corresponding to the test sample based on the comparison results.
[0038] Thirdly, embodiments of the present invention provide a computing device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the bearing fault classification method as described above.
[0039] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the bearing fault classification method as described above.
[0040] Fifthly, embodiments of the present invention provide a computer program product, the computer program product including a computer program, which, when executed by a processor, implements the bearing fault classification method as described above.
[0041] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following:
[0042] The bearing fault classification method provided in this invention constructs a sparse dictionary containing training samples of multiple fault types by processing first vibration signals under multiple known fault types. Test samples are obtained by processing acquired second vibration signals with fault classification. These test samples are then sparsely optimized to obtain first sparse coefficients. A reconstructed signal is obtained by reconstructing the signal using the first sparse coefficients and the training samples under multiple fault types. The fault type corresponding to the test sample is determined based on the comparison between the test sample and the reconstructed signal. Applying the method provided in this invention can accurately and effectively extract fault features from low signal-to-noise ratio vibration signals, greatly reducing the influence of limiting factors such as equipment operating environment, noise interference, and measurement point location. It can efficiently and accurately identify fault types in early bearing faults.
[0043] In the method for calculating the first sparse coefficient provided in this embodiment of the invention, the local features of the test sample are enhanced by calculating the RCMFE weighting parameters and weighting coefficients of the test sample, eliminating the calculation error caused by the time shift deviation of the test signal, fully considering the complexity of vibration signals of various fault types, and effectively improving the fault identification accuracy, especially when the early bearing fault characteristics are not obvious, effectively diagnosing and classifying bearing faults.
[0044] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings.
[0045] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0046] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:
[0047] Figure 1 A flowchart of a bearing fault classification method provided in an embodiment of the present invention;
[0048] Figure 2 A flowchart for constructing a sparse dictionary is provided for embodiments of the present invention;
[0049] Figure 3 A flowchart for calculating the first sparse coefficients provided in an embodiment of the present invention;
[0050] Figure 4 A flowchart of another bearing fault classification method provided in an embodiment of the present invention;
[0051] Figure 5 A schematic diagram illustrating the multi-sample test results provided in an embodiment of the present invention;
[0052] Figure 6 This is a structural block diagram of a bearing fault classification device provided in an embodiment of the present invention. Detailed Implementation
[0053] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
[0054] Now combined Figure 1 The steps of the bearing fault classification method in the embodiments of the present invention are described in detail.
[0055] Reference Figure 1 As shown in the figure, an embodiment of the present invention provides a bearing fault classification method, which includes the following steps:
[0056] S1. After processing the first vibration signal under multiple known fault types of the bearing, the collected signal is input into the preset dictionary learning algorithm model to construct a sparse dictionary containing training samples of multiple fault types.
[0057] S2. After processing the collected second vibration signal to be classified into faults, a test sample is obtained. The test sample is then subjected to sparse optimization solution to obtain the first sparse coefficient.
[0058] S3. Reconstruct the signals corresponding to the multiple fault types based on the first sparse coefficients and the training samples under multiple fault types respectively.
[0059] S4. Compare the test sample with the reconstruction signals corresponding to the multiple fault types, and determine the fault type corresponding to the test sample based on the comparison results.
[0060] Vibration signals from various known bearing fault states are collected using an accelerometer and referred to as the first vibration signal. The first vibration signal is then segmented and its amplitude is normalized.
[0061] Vibration signals of the bearing's fault condition to be classified are collected using an accelerometer and referred to as the second vibration signal. The amplitude of the second vibration signal is normalized to obtain the test sample.
[0062] In step S1 above, a sparse dictionary can be constructed, for example, in the following manner:
[0063] Reference Figure 2 As shown, the first vibration signal is input into the dictionary learning algorithm model to construct a sparse dictionary containing training samples of multiple fault types. This can be achieved in the following way:
[0064] S21. Establish a fixed initial sparse dictionary D;
[0065] S22. Using the improved K-SVD algorithm, the sparse coefficient matrix α can be obtained by the following formula:
[0066]
[0067] In the above formula, y is the sample matrix of the collected fault signals, D is the sparse dictionary, α is the sparse coefficient matrix, ||||0 is the l0 norm, that is, the number of non-zero elements in the vector, and ε is the sparse representation of the residual error.
[0068] S23. The signal is sparsely represented by the OMPerr algorithm, that is, the formula (1) is solved and the sparse matrix α is updated. During the update, α is divided into K parts and a column of α is updated separately each time to obtain the sparse coefficients:
[0069]
[0070] In the above formula, convergence occurs when the reconstruction error is less than ε, and the iteration stops; otherwise, the iteration continues.
[0071] The iterative process described above is also a process of updating the dictionary. The loss function for dictionary updates during this process can be expressed as follows:
[0072]
[0073] In the above formula, Y represents a matrix with k sample vectors, D represents a dictionary matrix, and A is a sparse coefficient matrix. Let d represent the j-th row of the sparse coefficient matrix A.k Let be the k-th atom of dictionary D, which is the update target. Updating dictionary D using SVD involves decomposing DA into the sum of k rank-1 matrices, and updating the k-th column d of dictionary D. k The corresponding k-th row in A is updated sequentially. During the update process, except for d, k and Repeat this process for all columns and rows except those in Y until all rows and columns are updated. Then merge all the fixed rows and columns with Y into E. k While SVD decomposition can indeed minimize the objective function, the resulting coefficient vector... Loss of sparsity, that is With update d k Before The non-zero elements are all in different positions and have different values, which means that the sparsity constraint is not satisfied.
[0074] To solve the above computational problem, we define... express The index of the non-zero element in the array, i.e. The index values of the points. Construct an N×ω... k The matrix Ω k , where the (ω)th k The values at (i),i) are non-zero, and all others are zero. Formula (3) can be written as:
[0075]
[0076] In the above formula, express Only retain row vectors with non-zero elements. Indicates from E k Remove the corresponding ω from the middle k The column obtained is non-zero.
[0077] In formula (4) The SVD decomposition yields the following expression:
[0078]
[0079] In the above formula, U is an m×m unitary matrix. VT is an m×n matrix whose diagonal elements are called singular values and are non-zero, and VT is an n×n unitary matrix.
[0080] Solve formula (5) and obtain the first column of matrix U and V through SVD decomposition. T The first line is respectively used as the updated d k and Repeat the update operation until the dictionary and sparse coefficients no longer change, then stop the iteration. The fully updated dictionary, i.e., the sparse dictionary D that can accurately match fault features, is obtained, as shown in the following equation:
[0081] D = [D1,D2,...,D] c ]∈R m′×n′ (6)
[0082] In the above formula, Let represent the i-th type of dictionary atom, C be the type of dictionary atom in the dictionary, and m′ and n′ represent the length and number of atoms in the dictionary, respectively.
[0083] Furthermore, referring to Figure 3 As shown, in step S2 above, the test samples are sparsely optimized to obtain the first sparse coefficients, which can be achieved through the following process:
[0084] S31. Calculate the sparsity coefficient of the test sample using formula (2), and call it the second sparsity coefficient;
[0085] Calculate the RECME weighted parameters for the test samples. Since the complexity of the vibration signals collected for various fault types differs, the composite multi-scale fuzzy entropy can effectively reflect the complexity of different time series. The formula is as follows:
[0086] MFE(X,m,n,r,N,τ)=FE(y i (τ),m,n,r,N) (7)
[0087] In the above formula, τ is the scaling factor, X is the time series, N is the length of X, and the embedding dimension m, similarity tolerance r, and gradient n are given in advance.
[0088] By fully considering the information in the time series, the average value of the series can be calculated, which is the composite multi-scale fuzzy entropy under that scale factor, as shown in the following formula:
[0089]
[0090] For time series with large scale factors, the fuzzy entropy value is calculated using the following formula to reduce the probability of undefined entropy values:
[0091]
[0092] The inventors discovered through research that RCMFE can express the interrelationships between time series and other time series at each scale factor, providing more comprehensive information than MFE and making the entropy values of different types of fault signals more distinct. During the calculation, the average value of the coarse-grained time series at each scale is calculated and used as the fuzzy entropy value for the corresponding scale, thereby reducing fluctuations caused by larger scale factors. At larger scale factors, the fuzzy entropy is calculated by summing the entropy values of all coarse-grained sequences at each scale factor, enhancing the adaptability of this method and reducing the probability of undefined entropy values. Therefore, RCMFE was chosen to calculate the weighting parameters.
[0093] S32 calculates the weighting coefficients based on the RCMFE weighting parameters, which can be achieved through the following steps:
[0094] The weighting coefficients are calculated using the following formula:
[0095] W(i)=A*|(hH(i)) n | (9)
[0096] In the above formula, h and H(i) represent the RCMFE weighting parameters of the test sample and the training sample under different fault types, respectively, A,n is a constant, and i is the fault type number. The weighting coefficients are calculated as follows:
[0097] w(i)=1 / W(i) (10)
[0098] S33. Based on the second sparsity coefficient and the weighting coefficient, perform sparsity optimization on the test samples to obtain the first sparsity coefficient, as shown in the following formula:
[0099]
[0100] in, ε is the first sparse coefficient, w is the weighting coefficient, D is the training sample in the sparse dictionary, and ε is the sparse representation residual error.
[0101] Furthermore, the first sparse coefficients and the training samples under multiple fault types are reconstructed respectively to obtain reconstructed signals corresponding to multiple fault types.
[0102] Reconstruction is the process of restoring the vibration signal from the sparse dictionary. When reconstructing the signal, test samples are typically used. The number of test samples is much smaller than the amount of data in the sparse dictionary. Based on the sparse dictionary and the test samples, relevant reconstruction algorithms can be applied to accurately reconstruct the vibration signal of the training samples. Specific reconstruction methods can be found in existing technologies; this embodiment of the invention does not limit these methods.
[0103] Furthermore, the test sample is compared with the reconstructed signals corresponding to the multiple fault types, and the fault type corresponding to the test sample is determined based on the comparison results. This can be achieved through the following steps:
[0104] The correlation between the test sample and the reconstructed signals corresponding to the multiple fault types is calculated using the following formula:
[0105]
[0106] In the above formula, N is the signal dimension, τ is the time shift deviation, and x(t) and y(t) are the signals with time shift deviations, respectively; R xy (t) represents the correlation between the test sample and the reconstructed signal;
[0107] The fault type corresponding to the reconstructed signal with the highest correlation is taken as the fault type of the test sample.
[0108] Let's take a specific example to illustrate the above-mentioned method for classifying bearing faults.
[0109] This example provides a detailed implementation process for the bearing fault classification method. (Refer to...) Figure 4 As shown, the specific implementation process of the bearing fault classification method is as follows:
[0110] 1. Collect the first vibration signal under multiple known fault types of the bearing and normalize it. Collect the second vibration signal to be classified into fault types and normalize it to obtain the test sample.
[0111] 2. Input the first vibration signal into the preset K-SVD dictionary learning algorithm model for learning;
[0112] 3. Obtain a sparse dictionary containing training samples of multiple fault types;
[0113] 4. Use OMPerr to solve for the second sparsity coefficients and calculate the RCMFE weighting parameters for the test samples;
[0114] 5. Solve for the weighting coefficients using the RCMFE weighting parameters;
[0115] 6. Based on the second sparsity coefficient and weighting coefficient of the test samples, perform weighted sparsity optimization to obtain the first sparsity coefficient;
[0116] 7. Reconstruct the reconstructed signal by combining the first sparse coefficients and the sparse dictionary of training samples for multiple fault types;
[0117] 8. Perform correlation analysis between the test samples and the reconstructed signals, and identify the fault type of the test samples based on the analysis results.
[0118] Reference Figure 5The diagram shows how the algorithm of this invention identifies four fault states for a cylindrical roller bearing of model NTN204: normal state, inner ring fault, outer ring fault, and rolling element fault. Multiple sample tests were conducted, with 50 test samples selected for each state (200 samples in total for the four states) for fault identification. Solid circles represent correctly identified fault types, and hollow circles represent incorrectly identified fault types. Experimental results are shown below. Figure 5 As shown, only 5 out of 200 test samples were incorrectly identified as fault types, while the remaining 195 samples were correctly classified, with a classification accuracy rate of 97.5%. This high accuracy rate proves that the present invention can be effectively applied to bearing fault diagnosis.
[0119] Based on the same inventive concept, this embodiment of the invention also provides a bearing failure classification device. Since the principle of solving the problem by these devices is similar to the aforementioned bearing failure classification method, the implementation of this device can refer to the implementation of the aforementioned method, and the repeated parts will not be described again.
[0120] This invention also provides a method and related apparatus for bearing fault classification, as described in the embodiments of the present invention. Figure 6 As shown, it includes:
[0121] The sparse dictionary construction module 61 is used to process the first vibration signal under multiple known fault types of the collected bearing and input it into the preset dictionary learning algorithm model to construct a sparse dictionary containing training samples of multiple fault types.
[0122] The first sparse coefficient solving module 62 is used to process the second vibration signal to be classified into faults to obtain a test sample, and to perform sparse optimization solving on the test sample to obtain the first sparse coefficient.
[0123] The signal reconstruction module 63 is used to reconstruct the signals based on the first sparse coefficients and training samples under multiple fault types to obtain reconstructed signals corresponding to multiple fault types.
[0124] The fault classification module 64 is used to compare the test sample with the reconstructed signals corresponding to the multiple fault types, and determine the fault type corresponding to the test sample based on the comparison result.
[0125] This invention provides a computing device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the bearing fault classification method as described above.
[0126] This invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the bearing fault classification method described above.
[0127] This invention provides a computer program product, which includes a computer program that, when executed by a processor, implements the bearing fault classification method described above.
[0128] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.
[0129] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can 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, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0130] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0131] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0132] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.
Claims
1. A method for classifying bearing faults, characterized in that, include: After processing the first vibration signal under multiple known fault types of the bearing, the collected signal is input into a preset dictionary learning algorithm model. The dictionary learning algorithm model is used to alternately calculate the sparse coefficients and update the dictionary until the dictionary and sparse coefficients no longer change. The iteration is stopped, and a sparse dictionary containing training samples of multiple fault types is obtained. After processing the second vibration signal to be classified for fault, test samples are obtained. The second sparsity coefficient and RCMFE weighting parameter of the test samples are calculated respectively. Calculate the weighting coefficients based on the RCMFE weighting parameters; Based on the second sparsity coefficient and the weighting coefficient, the test sample is sparsely optimized to obtain the first sparsity coefficient; Based on the first sparse coefficients and training samples under multiple fault types, reconstruction is performed to obtain reconstructed signals corresponding to multiple fault types. Determine the correlation between the test sample and the reconstructed signals corresponding to the multiple fault types, and determine the fault type corresponding to the test sample based on the correlation result; The step of calculating weighting coefficients based on the RCMFE weighting parameters, and performing sparse optimization on the test samples based on the second sparse coefficients and the weighting coefficients to obtain the first sparse coefficients includes: Calculate the following formula: ; In the above formula, and These represent the RCMFE weighting parameters for the test samples and the training samples under different fault types, respectively. is a constant, and i is the fault type number; The weighted coefficients are calculated as follows: ； The first sparsity coefficient is obtained by solving the sparsity optimization problem of the test samples using the following formula: ; in, Let w be the first sparse coefficient and w be the weighting coefficient. It is a sparse coefficient matrix. The sample matrix represents the collected fault signals, where D is the training sample in the sparse dictionary. The residual error is represented as sparse.
2. The method as described in claim 1, characterized in that, The correlation between the test sample and the reconstructed signals corresponding to the multiple fault types is determined by the following formula, and the fault type corresponding to the test sample is determined based on the result of the correlation: Calculate the correlation using the cross-correlation function formula: ; In the above formula, For signal dimension, This is due to time shift deviation. These are signals with time shift bias; To test the correlation between the samples and the reconstructed signal; The fault type corresponding to the reconstructed signal with the highest correlation is taken as the fault type of the test sample.
3. The method as described in claim 1 or 2, characterized in that, Processing the first vibration signal and processing the second vibration signal include: The first and second vibration signals are segmented and then normalized in amplitude.
4. A bearing failure classification device, characterized in that, include: The sparse dictionary construction module is used to process the first vibration signal under multiple known fault types of the collected bearing, and then input it into the preset dictionary learning algorithm model. The dictionary learning algorithm model is used to alternately calculate the sparse coefficients and update the dictionary until the dictionary and sparse coefficients no longer change, and then the iteration stops, resulting in a sparse dictionary containing training samples of multiple fault types. The first sparse coefficient solving module is used to process the acquired second vibration signal to be classified into test samples, calculate the second sparse coefficient and RCMFE weighting parameter of the test samples respectively, and calculate the weighting coefficient based on the RCMFE weighting parameter. Based on the second sparse coefficient and the weighting coefficient, sparse optimization is performed on the test samples to obtain the first sparse coefficient; the step of calculating the weighting coefficient based on the RCMFE weighting parameter, and performing sparse optimization on the test samples based on the second sparse coefficient and the weighting coefficient to obtain the first sparse coefficient includes: Calculate the following formula: ; In the above formula, and These represent the RCMFE weighting parameters for the test samples and the training samples under different fault types, respectively. Let i be a constant, and i be the fault type number. The weighting coefficients are calculated as follows: ； The first sparsity coefficient is obtained by solving the sparsity optimization problem of the test samples using the following formula: ; in, Let w be the first sparse coefficient and w be the weighting coefficient. It is a sparse coefficient matrix. The sample matrix represents the collected fault signals, where D is the training sample in the sparse dictionary. To represent residual error sparsely; The signal reconstruction module is used to reconstruct the signals based on the first sparse coefficients and training samples under multiple fault types to obtain reconstructed signals corresponding to multiple fault types. The fault classification module is used to determine the correlation between the test sample and the reconstructed signals corresponding to the multiple fault types, and to determine the fault type corresponding to the test sample based on the correlation result.
5. A computing device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the bearing fault classification method as described in any one of claims 1-3.
6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the bearing fault classification method as described in any one of claims 1 to 3.
7. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the bearing fault classification method as described in any one of claims 1 to 3.