Rolling bearing diagnosis method and system based on orthogonal projection mechanism and time domain orthogonal constraint

Abstract

The invention provides a rolling bearing diagnosis method and system based on an orthogonal projection mechanism and time domain orthogonal constraint. The method comprises the following steps: acquiring a vibration signal of a rolling bearing through a sensor; an orthogonal dictionary or an over-complete dictionary is adaptively selected according to the vibration signal intensity and frequency distribution characteristics, and dictionary parameter optimization and dictionary construction are completed according to a preset rule; based on the selected dictionary type, utilizing an ODMP or TOMP algorithm to calculate a sparse coefficient; reconstructing a signal based on the sparse coefficient; and analyzing an envelope spectrum of the reconstructed signal to extract a fault characteristic frequency and a harmonic wave thereof, modulating a sideband, and completing rolling bearing fault type determination. According to the method, pseudo-inverse calculation and repeated projection operation of a traditional OMP algorithm are eliminated through an orthogonal projection mechanism and time domain orthogonal constraint, the calculation complexity of the ODMP algorithm is reduced to be linear, the complexity of the TOMP algorithm is remarkably reduced, the accuracy of sparse coefficient estimation is ensured, and the operation time is greatly shortened; the method is suitable for rolling bearing early weak fault detection.

Description

Technical Field

[0001] This invention relates to the field of rolling bearing fault diagnosis technology, and in particular to a rolling bearing diagnosis method and system based on orthogonal projection mechanism and time-domain orthogonal constraints. Background Technology

[0002] Rolling bearings, as core components of rotating machinery, are widely used in machine tools, motors, rail transportation, aerospace, and other fields. Their operating condition directly determines the reliability, safety, and service life of the equipment. Throughout the rolling bearing's lifecycle, early-stage faults are easily masked by background noise and harmonic interference generated during equipment operation due to their weak impact energy, making them difficult to identify directly. However, if not diagnosed and addressed promptly, these early faults can rapidly develop into serious malfunctions, leading to equipment downtime, production interruptions, and even safety accidents. Therefore, achieving rapid and accurate diagnosis of early-stage rolling bearing faults is of significant engineering importance.

[0003] Sparse representation techniques, with their powerful noise suppression and feature extraction capabilities, have become one of the mainstream techniques for reconstructing the impact components of rolling bearing faults. The Orthogonal Matching Pursuit (OMP) algorithm, due to its flexible implementation and excellent sparse recovery performance, is widely used for solving sparse coefficients. However, traditional OMP algorithms and existing improved variants suffer from the following core defects, limiting their engineering applications: 1) Low computational efficiency: Traditional OMP algorithms require pseudo-inverse calculations (such as SVD decomposition) and repeated inner product projection operations in each iteration. The computational complexity increases quadratically with the signal length L and dictionary size N, making it difficult to handle long-term vibration signals or large-scale data. 2) Incomplete utilization of dictionary and signal characteristics: For orthogonal dictionaries (such as DCT and Fourier basis), OMP algorithms do not utilize atomic orthogonality, leading to redundant calculations. For overcomplete dictionaries (such as wavelet dictionaries), they do not combine the sparse and non-overlapping physical characteristics of fault impact in the time domain, still relying on complex iterative solutions. 3) Insufficient real-time performance: High computational costs make it difficult to deploy traditional OMP algorithms and improved variants in industrial real-time monitoring scenarios, especially unsuitable for online diagnosis of high-speed bearings (such as train wheelset bearings and aero-engine bearings). 4) Reliance on prior knowledge or manual intervention: Some improved solutions require pre-construction of fault models, setting of empirical parameters, or reliance on dictionary learning training processes, resulting in poor adaptability and unsuitability for complex and variable working conditions.

[0004] Therefore, there is an urgent need to develop a rolling bearing diagnostic method and system that can fully utilize the characteristics of dictionary structure and signal physical properties, significantly reduce computational complexity while maintaining diagnostic accuracy, and possess strong adaptability and engineering practicality. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a rolling bearing diagnosis method and system based on orthogonal projection mechanism and temporal orthogonal constraints. This invention designs optimization algorithms for orthogonal dictionaries and overcomplete dictionaries respectively. By introducing orthogonal projection mechanism and temporal orthogonal constraints, redundant calculations of traditional OMP algorithm are fundamentally eliminated, significantly improving computational efficiency while ensuring diagnostic accuracy, and realizing rapid and accurate diagnosis of rolling bearing faults.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a rolling bearing diagnostic method based on orthogonal projection mechanism and temporal orthogonal constraints, comprising the following steps: S1) The vibration signal of the rolling bearing is collected by a sensor, and the vibration signal includes fault impact components and background noise; S2) Based on the vibration signal intensity and frequency distribution characteristics, an orthogonal dictionary or an overcomplete dictionary is adaptively selected, and dictionary parameter optimization and dictionary construction are completed according to preset rules; S3) Calculate the sparsity coefficients based on the selected dictionary type; If an orthogonal dictionary is chosen, the orthogonal dictionary matching pursuit ODMP algorithm based on the orthogonal projection mechanism is adopted, and the sparsity coefficients are calculated by a single inner product projection using atomic orthogonality. If an overcomplete dictionary is selected, the TOMP algorithm based on time-domain orthogonal matching pursuit is adopted. The time non-overlapping property of fault impact forces the selected atoms to be orthogonal, and the sparsity coefficients are calculated. S4) Reconstruct the signal based on the sparse coefficients obtained from the solution; S5) Perform envelope spectrum analysis on the reconstructed signal to extract the fault characteristic frequency FCF, harmonic components and modulation sidebands, compare them with the preset rolling bearing fault frequency database, and output the fault type, fault characteristic parameters and diagnostic confidence level to realize rolling bearing fault diagnosis.

[0007] Preferably, in step S2), the orthogonal dictionary... for: In the formula, Indicates the first An orthogonal dictionary atom, It is determined by the sampling rate and vibration signal length The determined frequency resolution; Number of atoms; The Nyquist frequency was determined based on the sampling rate, and the number of atoms constructed was... An orthogonal dictionary, for an orthogonal dictionary Any two different atoms and Its inner volume .

[0008] Preferably, in step S2), the orthogonal dictionary includes a discrete cosine DCT dictionary, a Fourier basis dictionary, and an orthogonal wavelet dictionary.

[0009] Preferably, in step S2), the overcomplete dictionary includes a Laplace wavelet dictionary, a Moray wavelet dictionary, and a harmonic wavelet dictionary.

[0010] Preferably, in step S3), the orthogonal dictionary matching pursuit (ODMP) algorithm based on orthogonal projection mechanism is used to calculate the sparsity coefficients through a single inner product projection using atomic orthogonality. Specifically, it includes the following steps: S311) Input vibration signal Orthogonal dictionaries that satisfy atomic orthogonality and preset frequency range, initial coefficient set and set stopping criteria; S312) Calculate vibration signal Orthogonal dictionary Correlation of atoms in the middle ; S313) If the stopping criterion is a fixed number of iterations Select the one with the largest correlation magnitude. The set of indices corresponding to each atom Directly determine the sparsity coefficient ; S314) If the stopping criterion is the target residual, then calculate the coefficients of the selected atoms. And calculate the residual signal. ; S315) Determine whether the L2 norm of the residual signal is less than the threshold. That is, to judge If the condition is met, stop the iteration; otherwise, repeat steps S23-S24. S316), Output sparsity coefficient and residual signals It is used for signal reconstruction.

[0011] Preferably, in step S3), the time-domain orthogonal matching pursuit (TOMP) algorithm based on time-domain orthogonality constraints is used to force the selected atoms to be orthogonal by utilizing the time non-overlapping nature of fault impacts. Specifically, this includes: S321) Input vibration signal Complete dictionary Atom compact support length Initialize the index set Initial residual signal Number of iterations ; S322) The optimal parameters of the overcomplete dictionary are determined by the correlation filter (CF) method, i.e.: A candidate atom set is constructed within a preset parameter range, and the vibration signal is calculated. The normalized correlation coefficient with each candidate atom is used to select the parameter corresponding to the candidate atom with the highest correlation coefficient as the optimal parameter; S323) Calculate the current residual With complete dictionary The correlation vector of all atoms in ; S324) When the stopping condition is not met, filter the relevance vector. Atom index with the largest amplitude index Add to index set ,renew ; S225), eliminating correlation within the neighborhood, atoms The correlation vector elements within the neighborhood range are set to 0, forcing the selected atom to remain orthogonal to its neighboring atoms, i.e.: in, Indicates the length of the compact support; S326) Determine if the stopping criterion is met; if so, stop the iteration and output the support set. I sparsity coefficient .

[0012] Preferably, in step S322), the parameters include atomic frequency parameters. Damping ratio In the aforementioned preset parameter range, the atomic frequency parameter Conforms to the Nyquist sampling criterion, damping ratio The range is limited to [0, 0.2], and the atomic compactness supports a length of [length]. Determined based on the minimum time interval of the fault impact.

[0013] Preferably, the time non-overlapping property of the failure impact is such that the time interval between rolling bearing failure impact events is greater than the atomic compact support length. This ensures that the support sets of any two selected atoms are mutually exclusive.

[0014] Preferably, in step S4), the expression for reconstructing the vibration signal using the sparse coefficients of the orthogonal dictionary matching tracing ODMP algorithm is as follows: ; In the formula, The first in the orthogonal dictionary One atom; The expression for reconstructing the vibration signal using the sparse coefficients of the TOMP algorithm in the time domain is as follows: In the formula, For the reconstructed vibration signal, Represents an overcomplete dictionary set; Represents the set of sparse coefficients; Indicates the first The iteration of the ... The sparsity coefficient of each atom; The first character of a complete dictionary One atom.

[0015] Preferably, the method further includes atomic adaptive matching, that is, dynamically adjusting the atomic support length based on waveform similarity and minimum impact interval, wherein waveform similarity is quantified and evaluated by normalized correlation coefficient, and minimum impact interval is derived from the fault characteristic frequency FCF in the vibration signal, so as to achieve accurate matching between atomic shape and fault impact characteristics.

[0016] Secondly, the present invention provides a rolling bearing diagnostic system based on orthogonal projection mechanism and time-domain orthogonal constraints, comprising: The signal acquisition module acquires the vibration signal of the rolling bearing through a sensor. The vibration signal includes fault impact components and background noise. The dictionary adaptation construction module analyzes the noise intensity and frequency distribution characteristics of the preprocessed signal output by the signal acquisition module, adaptively selects an orthogonal dictionary or an overcomplete dictionary, and completes dictionary parameter optimization and dictionary construction according to preset rules. The sparse coefficient solving module has built-in Orthogonal Dictionary Matching Pursuit (ODMP) and Temporal Orthogonal Matching Pursuit (TOMP) algorithms. It automatically calls the corresponding algorithm to solve the sparse coefficients based on the dictionary type output by the dictionary adaptation module. The signal reconstruction module, based on the sparse coefficients output by the sparse coefficient solving module, combines the dictionary output by the dictionary adaptation module to complete the vibration signal reconstruction. The fault intelligent identification module performs envelope spectrum analysis on the reconstructed signal, extracts the fault characteristic frequency FCF, harmonic components and modulation sidebands, compares them with the preset rolling bearing fault frequency database, and outputs the fault type, fault characteristic parameters and diagnostic confidence level.

[0017] The beneficial effects of this invention are as follows: 1. This invention completely eliminates the pseudo-inverse calculation and repeated projection operation of the traditional OMP algorithm through the orthogonal projection mechanism and time-domain orthogonal constraints. The computational complexity of the ODMP algorithm is reduced to linear, the complexity of the TOMP algorithm is significantly reduced, and the running time is shortened by two orders of magnitude. It can meet the computational efficiency requirements of real-time monitoring in industrial sites and is suitable for high-speed and long-term signal diagnosis scenarios. 2. This invention ensures the accuracy of sparse coefficient estimation through orthogonality constraints and adaptive atom matching mechanism, avoiding the loss of fault features. Even if the fault impact energy is weak, the fault feature frequency and its harmonics and modulation sidebands can still be clearly extracted, with a diagnostic accuracy of 100%. It is suitable for early weak fault detection in rolling bearings. 3. This invention optimizes dictionary parameters through correlation filtering and adaptively adjusts the atomic support length. It does not require pre-constructing fault models or manually setting empirical parameters. It can operate stably under different working conditions, different fault types, and strong noise environments, reducing the dependence on the professional skills of operators. 4. This invention is compatible with orthogonal dictionaries and overcomplete dictionaries, and is suitable for fault diagnosis of various types of rolling bearings. It can be widely used in rolling bearing condition monitoring and fault diagnosis in machine tools, motors, rail transportation, aerospace and other fields. Attached Figure Description

[0018] Figure 1 This is a schematic flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the orthogonal dictionary matching tracing (ODMP) algorithm of this invention. Figure 3 This is a flowchart illustrating the temporal orthogonal matching pursuit (TOMP) algorithm of the present invention. Figure 4 This is a comparison chart of the running time of the ODMP of this invention and the traditional OMP under different parameters; Figure 5 The present invention relates to the bearing fault vibration signal and its envelope spectrum at a speed of 100 km / h for the train bearing fault signal. Figure 6 This is a graph showing the ODMP analysis results in an embodiment of the present invention; Figure 7 This is a simulation result diagram of the bearing outer ring fault signal in an embodiment of the present invention; Figure 8 This is a schematic diagram of the analog signal reconstructed using the TOMP algorithm in an embodiment of the present invention; Figure 9 This is a schematic diagram of the outer ring fault signal in an embodiment of the present invention; Figure 10 This is a diagram showing the analysis results of the bearing outer ring fault signal by TOMP in an embodiment of the present invention; Figure 11 This is a schematic diagram of the inner loop fault signal in an embodiment of the present invention; Figure 12 This is a diagram showing the analysis results of the bearing inner ring signal using TOMP in an embodiment of the present invention. Detailed Implementation

[0019] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings: Example 1 like Figure 1 As shown, this embodiment provides a rolling bearing diagnostic method based on orthogonal projection mechanism and temporal orthogonal constraints, including the following steps: S1) The vibration signal of the rolling bearing is collected by a sensor, and the vibration signal includes fault impact components and background noise; S2) Based on the vibration signal intensity and frequency distribution characteristics, an orthogonal dictionary or an overcomplete dictionary is adaptively selected, and dictionary parameter optimization and dictionary construction are completed according to preset rules; In this embodiment, the orthogonal dictionary for: In the formula, Indicates the first An orthogonal dictionary atom, It is determined by the sampling rate and vibration signal length The determined frequency resolution; This represents the number of atoms.

[0020] In this embodiment, the Nyquist frequency is determined based on the sampling frequency, and the number of atoms constructed is... An orthogonal dictionary, and for an orthogonal dictionary Any two different atoms and Its inner volume .

[0021] The orthogonal dictionaries include Discrete Cosine Transform (DCT) dictionaries, Fourier basis dictionaries, and orthogonal wavelet dictionaries.

[0022] The comprehensive dictionaries include the Laplace wavelet dictionary, the Molay wavelet dictionary, and the harmonic wavelet dictionary.

[0023] S3) Calculate the sparsity coefficients based on the selected dictionary type; If an orthogonal dictionary is chosen, the orthogonal dictionary matching pursuit ODMP algorithm based on the orthogonal projection mechanism is adopted, and the sparsity coefficients are calculated by a single inner product projection using atomic orthogonality. If an overcomplete dictionary is selected, the TOMP algorithm based on time-domain orthogonal matching pursuit is adopted. The time non-overlapping property of fault impact forces the selected atoms to be orthogonal, and the sparsity coefficients are calculated. In this embodiment, the Orthogonal Dictionary Matching Pursuit (ODMP) algorithm based on orthogonal projection mechanism is adopted. It utilizes atomic orthogonality to calculate sparse coefficients through a single inner product projection, specifically including the following steps: S311) Input vibration signal Orthogonal dictionaries that satisfy atomic orthogonality and preset frequency range, initial coefficient set and set stopping criteria; S312) Calculate vibration signal Orthogonal dictionary Correlation of atoms in the middle ; S313) If the stopping criterion is a fixed number of iterations Select the one with the largest correlation magnitude. The set of indices corresponding to each atom Directly determine the sparsity coefficient ; S314) If the stopping criterion is the target residual, then calculate the coefficients of the selected atoms. And calculate the residual signal. ; S315) Determine whether the L2 norm of the residual signal is less than the threshold. That is, to judge If the condition is met, stop the iteration; otherwise, repeat steps S23-S24. S316), Output sparsity coefficient and residual signals It is used for signal reconstruction.

[0024] In this embodiment, the temporal orthogonal matching pursuit (TOMP) algorithm based on temporal orthogonal constraints is adopted. This algorithm utilizes the temporal non-overlapping nature of fault impacts to force the selected atoms to be orthogonal. Specifically, it includes: S321) Input vibration signal Complete dictionary Atom compact support length Initialize the index set Initial residual signal Number of iterations ; S322) The optimal parameters of the overcomplete dictionary are determined by the correlation filter (CF) method, i.e.: A candidate atom set is constructed within a preset parameter range, and the vibration signal is calculated. The normalized correlation coefficient with each candidate atom is used to select the parameter corresponding to the candidate atom with the highest correlation coefficient as the optimal parameter; The parameters mentioned include atomic frequency parameters. Damping ratio In the aforementioned preset parameter range, the atomic frequency parameter Conforms to the Nyquist sampling criterion, damping ratio The range is limited to [0, 0.2], and the atomic compactness supports a length of [length]. Determined based on the minimum time interval of the fault impact.

[0025] S323) Calculate the current residual With complete dictionary The correlation vector of all atoms in ; S324) When the stopping condition is not met, filter the relevance vector. Atom index with the largest amplitude index Add to index set ,renew ; S325), eliminating correlation within the neighborhood, atoms The correlation vector elements within the neighborhood range are set to 0, forcing the selected atom to remain orthogonal to its neighboring atoms, i.e.: in, Indicates the length of the compact support; S326) Determine if the stopping criterion is met; if so, stop the iteration and output the support set. I sparsity coefficient .

[0026] In this embodiment, the temporal non-overlapping property of the fault impact is defined as follows: the time interval between rolling bearing fault impact events is greater than the atomic compact support length. This ensures that the support sets of any two selected atoms are mutually exclusive.

[0027] S4) Reconstruct the signal based on the sparse coefficients obtained from the solution; The expression for reconstructing the vibration signal using the sparse coefficients of the ODMP algorithm with orthogonal dictionary matching is as follows: ; In the formula, The first in the orthogonal dictionary One atom; The expression for reconstructing the vibration signal using the sparse coefficients of the TOMP algorithm in the time domain is as follows: In the formula, For the reconstructed vibration signal, Represents an overcomplete dictionary set; Represents the set of sparse coefficients; Indicates the first The iteration of the ... The sparsity coefficient of each atom; The first character of a complete dictionary One atom.

[0028] S5) Perform envelope spectrum analysis on the reconstructed signal to extract the fault characteristic frequency FCF, harmonic components and modulation sidebands, compare them with the preset rolling bearing fault frequency database, and output the fault type, fault characteristic parameters and diagnostic confidence level to realize rolling bearing fault diagnosis.

[0029] In this embodiment, the method further includes atomic adaptive matching, namely: dynamically adjusting the atomic support length based on waveform similarity and minimum impact interval, wherein waveform similarity is quantified and evaluated by normalized correlation coefficient, and the minimum impact interval is derived from the fault characteristic frequency FCF in the vibration signal, thereby achieving accurate matching between atomic shape and fault impact characteristics.

[0030] like Figure 4 As shown, this embodiment uses MATLAB installed on a desktop computer equipped with a 3.50GHz CPU and 32GB RAM to perform signal reconstruction using ODMP. A traditional OMP algorithm is used for comparison. In this embodiment, the ODMP algorithm based on the stopping criterion of iteration count is denoted as ODMP1, and the algorithm based on the stopping criterion of target error is denoted as ODMP2. The running times of the OMP, ODMP1, and ODMP2 algorithms are 8.80 seconds, 0.17 seconds, and 0.24 seconds, respectively. MATLAB performance analysis results show that the matrix transpose and inner product operations in the OMP algorithm take as long as 8.20 seconds, while the pseudo-inverse matrix calculation only takes 0.36 seconds. To further evaluate performance, this embodiment tests the algorithm running time under different parameter settings, such as... Figure 4 As shown in the figure, the results indicate that the computation time of the OMP algorithm is significantly affected by the signal length, dictionary size, and number of iterations. In contrast, the time consumption of ODMP2 is mainly affected by the number of iterations, but the overall computational cost remains at a low level; while ODMP1 is almost insensitive to the number of iterations and always maintains the lowest running time.

[0031] The ODMP algorithm can significantly reduce computation time. Compared to ODMP1, ODMP2 incurs some additional overhead because it needs to calculate residuals to determine the stopping criterion. It is worth noting that if the stopping criterion of the OMP algorithm is set to be the same as that of ODMP2, this additional overhead is negligible compared to the time-consuming repeated projections and pseudo-inverse calculations during the iteration process of OMP, and therefore can be ignored in the performance evaluation.

[0032] Figure 5 This invention presents a bearing fault vibration signal and its envelope spectrum at a speed of 100 km / h, illustrating a train bearing fault signal. Identifying fault characteristic frequencies from the envelope spectrum is challenging. The ODMP algorithm is employed to suppress harmonic interference.

[0033] Figure 6 The results of the ODMP algorithm are shown. As can be seen from the figure, the fault characteristic frequencies and their harmonics are obvious in its envelope spectrum, and the fault characteristics are revealed.

[0034] This embodiment constructs a simulated bearing outer ring fault signal for analysis, such as... Figure 7 As shown, the results obtained using the TOMP algorithm in this embodiment are displayed in... Figure 8 In this embodiment, the computational costs of the complete process (including parameter identification, dictionary construction, and coefficient solving (TOMP)) are 9.35 seconds, 3.44 seconds, and 0.14 seconds, respectively, demonstrating the efficiency of the overall method.

[0035] Figure 9 The time-domain waveform and envelope spectrum of the outer ring fault vibration signal are shown. The fault pulse is completely masked by significant interference, and no fault-related features are shown in the envelope spectrum. In this embodiment, the atomic frequency parameters are first determined using the correlation filtering method. Damping ratio The iteration count K is set to approximately twice the estimated number of fault pulses. The TOMP reconstructed signal results are as follows: Figure 10 As shown.

[0036] Figure 11 In this embodiment, the inner ring fault vibration signal is suppressed by strong noise in the time domain, and the envelope spectrum does not show obvious fault characteristics. This embodiment uses the TOMP method to reconstruct the inner ring fault signal, and the atomic frequency parameters are used. =7290Hz, damping ratio =0.07, determined using correlation filtering, with iteration number K=260. The result is as follows... Figure 12 As shown, the envelope spectrum reveals the bearing fault frequency (FCF), its harmonics, and the modulation sidebands associated with the bearing rotation frequency.

[0037] Example 2 This embodiment provides a rolling bearing diagnostic system based on orthogonal projection mechanism and time-domain orthogonal constraints, including: The signal acquisition module acquires the vibration signal of the rolling bearing through a sensor. The vibration signal includes fault impact components and background noise. In this embodiment, the signal acquisition module consists of an accelerometer, a data acquisition card, and a signal preprocessing unit. The accelerometer is installed in the bearing housing or near the machine body to collect the vibration signal of the rolling bearing; the data acquisition card converts the analog signal into a digital signal; and the signal preprocessing unit performs filtering, de-trending, and other operations to remove low-frequency drift and high-frequency interference from the signal.

[0038] The dictionary adaptation construction module analyzes the noise intensity and frequency distribution characteristics of the preprocessed signal output by the signal acquisition module, adaptively selects an orthogonal dictionary or an overcomplete dictionary, and completes dictionary parameter optimization and dictionary construction according to preset rules. The sparse coefficient solving module has built-in Orthogonal Dictionary Matching Pursuit (ODMP) and Temporal Orthogonal Matching Pursuit (TOMP) algorithms. It automatically calls the corresponding algorithm to solve the sparse coefficients based on the dictionary type output by the dictionary adaptation module. The signal reconstruction module, based on the sparse coefficients output by the sparse coefficient solving module, combines the dictionary output by the dictionary adaptation module to complete the vibration signal reconstruction. The fault intelligent identification module performs envelope spectrum analysis on the reconstructed signal, extracts the fault characteristic frequency FCF, harmonic components and modulation sidebands, compares them with the preset rolling bearing fault frequency database, and outputs the fault type, fault characteristic parameters and diagnostic confidence level.

[0039] The embodiments and descriptions above are merely illustrative of the principles and preferred embodiments of the present invention. Various changes and modifications may be made to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed.

Claims

1. A method for diagnosing rolling bearings based on orthogonal projection mechanism and temporal orthogonal constraints, characterized in that, Includes the following steps: S1) The vibration signal of the rolling bearing is collected by a sensor, and the vibration signal includes fault impact components and background noise; S2) Based on the vibration signal intensity and frequency distribution characteristics, an orthogonal dictionary or an overcomplete dictionary is adaptively selected, and dictionary parameter optimization and dictionary construction are completed according to preset rules; S3) Calculate the sparsity coefficients based on the selected dictionary type; If an orthogonal dictionary is chosen, the orthogonal dictionary matching pursuit ODMP algorithm based on the orthogonal projection mechanism is adopted, and the sparsity coefficients are calculated by a single inner product projection using atomic orthogonality. If an overcomplete dictionary is selected, the TOMP algorithm based on time-domain orthogonal matching pursuit is adopted. The time non-overlapping property of fault impact forces the selected atoms to be orthogonal, and the sparsity coefficients are calculated. S4) Reconstruct the vibration signal based on the sparse coefficients obtained from the solution; S5) Perform envelope spectrum analysis on the reconstructed signal to extract the fault characteristic frequency FCF, harmonic components and modulation sidebands, compare them with the preset rolling bearing fault frequency database, and output the fault type, fault characteristic parameters and diagnostic confidence level to realize rolling bearing fault diagnosis.

2. The rolling bearing diagnostic method based on orthogonal projection mechanism and temporal orthogonal constraints according to claim 1, characterized in that: In step S2), the orthogonal dictionary for: In the formula, Indicates the first An orthogonal dictionary atom, It is determined by the sampling rate and vibration signal length The determined frequency resolution; Number of atoms; The Nyquist frequency was determined based on the sampling rate, and the number of atoms constructed was... An orthogonal dictionary, for an orthogonal dictionary Any two different atoms and Its inner volume .

3. The rolling bearing diagnostic method based on orthogonal projection mechanism and temporal orthogonal constraints according to claim 2, characterized in that: In step S2), the orthogonal dictionary includes a discrete cosine DCT dictionary, a Fourier basis dictionary, and an orthogonal wavelet dictionary.

4. The rolling bearing diagnostic method based on orthogonal projection mechanism and temporal orthogonal constraints according to claim 1, characterized in that: In step S2), the overcomplete dictionary includes a Laplace wavelet dictionary, a Molay wavelet dictionary, and a harmonic wavelet dictionary.

5. The rolling bearing diagnostic method based on orthogonal projection mechanism and temporal orthogonal constraints according to claim 1, characterized in that: In step S3), the orthogonal dictionary matching pursuit (ODMP) algorithm based on orthogonal projection mechanism is adopted. The sparsity coefficients are calculated by single inner product projection using atomic orthogonality. Specifically, the steps are as follows: S311) Input vibration signal Orthogonal dictionaries that satisfy atomic orthogonality and preset frequency range, initial coefficient set and set stopping criteria; S312) Calculate vibration signal Orthogonal dictionary Correlation of atoms in ; S313) If the stopping criterion is a fixed number of iterations Select the one with the largest correlation magnitude. The set of indices corresponding to each atom Directly determine the sparsity coefficient ; S314) If the stopping criterion is the target residual, then calculate the coefficients of the selected atoms. And calculate the residual signal. ; S315) Determine whether the L2 norm of the residual signal is less than the threshold. That is, to judge If the condition is met, stop the iteration; otherwise, repeat steps S23-S24. S316), Output sparsity coefficient and residual signals It is used for signal reconstruction.

6. The rolling bearing diagnostic method based on orthogonal projection mechanism and temporal orthogonal constraints according to claim 1, characterized in that: In step S3), the TOMP algorithm based on time-domain orthogonal matching pursuit, which utilizes the time non-overlapping nature of fault impacts to force the selected atoms to be orthogonal, specifically includes: S321) Input vibration signal Complete dictionary Atom compact support length Initialize the index set Initial residual signal Number of iterations ; S322) The optimal parameters of the overcomplete dictionary are determined by the correlation filter (CF) method, i.e.: A candidate atom set is constructed within a preset parameter range, and the vibration signal is calculated. The normalized correlation coefficient with each candidate atom is used to select the parameter corresponding to the candidate atom with the highest correlation coefficient as the optimal parameter; S323) Calculate the current residual With complete dictionary The correlation vector of all atoms in ; S324) When the stopping condition is not met, filter the relevance vector. Atom index with the largest amplitude index Add to index set ,renew ; S325), eliminating correlation within the neighborhood, atoms The correlation vector elements within the neighborhood range are set to 0, forcing the selected atom to remain orthogonal to its neighboring atoms, i.e.: in, Indicates the length of the compact support; S326) Determine if the stopping criterion is met; if so, stop the iteration and output the support set. I sparsity coefficient .

7. The rolling bearing diagnostic method based on orthogonal projection mechanism and temporal orthogonal constraints according to claim 6, characterized in that: In step S322), the parameters include atomic frequency parameters. Damping ratio In the aforementioned preset parameter range, the atomic frequency parameter Conforms to the Nyquist sampling criterion, damping ratio The range is limited to [0, 0.2], and the atomic compactness supports a length of [length]. Determined based on the minimum time interval of the fault impact.

8. The rolling bearing diagnostic method based on orthogonal projection mechanism and temporal orthogonal constraints according to claim 7, characterized in that: The aforementioned non-overlapping time of failure impacts is defined as follows: the time interval between rolling bearing failure impact events is greater than the atomic compact support length. This ensures that the support sets of any two selected atoms are mutually exclusive.

9. The rolling bearing diagnostic method based on orthogonal projection mechanism and temporal orthogonal constraints according to claim 8, characterized in that: In step S4), the expression for reconstructing the vibration signal using the sparse coefficients of the orthogonal dictionary matching tracing ODMP algorithm is as follows: ； In the formula, The first in the orthogonal dictionary One atom; The expression for reconstructing the vibration signal using the sparse coefficients of the TOMP algorithm in the time domain is as follows: In the formula, For the reconstructed vibration signal, Represents an overcomplete dictionary set; Represents the set of sparse coefficients; Indicates the first The iteration of the ... The sparsity coefficient of each atom; The first character of a complete dictionary One atom.

10. A rolling bearing diagnostic system based on orthogonal projection mechanism and time-domain orthogonal constraints, characterized in that, include: The signal acquisition module acquires the vibration signal of the rolling bearing through a sensor. The vibration signal includes fault impact components and background noise. The dictionary adaptation construction module analyzes the noise intensity and frequency distribution characteristics of the preprocessed signal output by the signal acquisition module, adaptively selects an orthogonal dictionary or an overcomplete dictionary, and completes dictionary parameter optimization and dictionary construction according to preset rules. The sparse coefficient solving module has built-in Orthogonal Dictionary Matching Pursuit (ODMP) and Temporal Orthogonal Matching Pursuit (TOMP) algorithms. It automatically calls the corresponding algorithm to solve the sparse coefficients based on the dictionary type output by the dictionary adaptation module. The signal reconstruction module, based on the sparse coefficients output by the sparse coefficient solving module, combines the dictionary output by the dictionary adaptation module to complete the vibration signal reconstruction. The fault intelligent identification module performs envelope spectrum analysis on the reconstructed signal, extracts the fault characteristic frequency FCF, harmonic components and modulation sidebands, compares them with the preset rolling bearing fault frequency database, and outputs the fault type, fault characteristic parameters and diagnostic confidence level.

Images