CN-122016314-A - Rolling bearing diagnosis method and system based on orthogonal projection mechanism and time domain orthogonal constraint

CN122016314ACN 122016314 ACN122016314 ACN 122016314ACN-122016314-A

Abstract

The invention provides a rolling bearing diagnosis method and a system based on orthogonal projection mechanism and time domain orthogonal constraint, wherein a sensor is used for collecting vibration signals of a rolling bearing; the method comprises the steps of adaptively selecting an orthogonal dictionary or an overcomplete dictionary according to the intensity and frequency distribution characteristics of vibration signals, completing dictionary parameter optimization and dictionary construction according to preset rules, calculating sparse coefficients by utilizing ODMP or TOMP algorithm based on the selected dictionary types, reconstructing signals based on the sparse coefficients, extracting fault characteristic frequencies, harmonic waves and modulation sidebands of the fault characteristic frequencies by analyzing envelope spectrums of the reconstructed signals, and completing rolling bearing fault type judgment. According to the invention, through an orthogonal projection mechanism and time domain orthogonal constraint, the pseudo-inverse calculation and repeated projection operation of the traditional OMP algorithm are eliminated, the calculation complexity of the ODMP algorithm is reduced to be linear, the complexity of the TOMP algorithm is obviously reduced, the accuracy of sparse coefficient estimation is ensured, the running time is greatly shortened, and the method is suitable for early weak fault detection of the rolling bearing.

Inventors

LI YONGJIAN
WU ZHANHUA
WU YUYUAN
GUO XINYUAN
LIANG HAIYING
LIU RUNHAO

Assignees

五邑大学

Dates

Publication Date: 20260512
Application Date: 20260106

Claims (10)

1. A rolling bearing diagnosis method based on orthogonal projection mechanism and time domain orthogonal constraint is characterized by comprising the following steps: s1), collecting vibration signals of a rolling bearing through a sensor, wherein the vibration signals comprise fault impact components and background noise; s2), adaptively selecting an orthogonal dictionary or an overcomplete dictionary according to the intensity and frequency distribution characteristics of the vibration signals, and completing dictionary parameter optimization and dictionary construction according to preset rules; s3), calculating sparse coefficients based on the selected dictionary types; If an orthogonal dictionary is selected, adopting an orthogonal dictionary matching pursuit ODMP algorithm based on an orthogonal projection mechanism, and calculating a sparse coefficient through single inner product projection by utilizing atom orthogonality; If the overcomplete dictionary is selected, adopting a time domain orthogonal matching pursuit TOMP algorithm based on time domain orthogonal constraint, forcing the selected atoms to be orthogonal by utilizing the time non-overlapping property of fault impact, and calculating a sparse coefficient; s4), reconstructing a vibration signal based on the sparse coefficient obtained by solving; s5), carrying out envelope spectrum analysis on the reconstructed signal, extracting fault characteristic frequency FCF, harmonic component and modulation sideband, comparing with a preset rolling bearing fault frequency database, outputting fault type, fault characteristic parameter and diagnosis confidence, and realizing rolling bearing fault diagnosis.
2. The method for diagnosing a rolling bearing based on an orthogonal projection mechanism and time domain orthogonal constraint according to claim 1, wherein in step S2), the orthogonal dictionary is used for The method comprises the following steps: In the formula, Represent the first The number of orthogonal dictionary atoms is one, Is made up of sampling rate And vibration signal length A determined frequency resolution; Is the number of atoms; Determining Nyquist frequency based on sampling rate, and constructing atomic number as For orthogonal dictionary of (2) Any two different atoms And Its inner product 。
3. The method for diagnosing a rolling bearing based on an orthogonal projection mechanism and time domain orthogonal constraint according to claim 2, wherein in the step S2), the orthogonal dictionary comprises a discrete cosine DCT dictionary, a Fourier base dictionary and an orthogonal wavelet dictionary.
4. The method for diagnosing a rolling bearing based on an orthogonal projection mechanism and time domain orthogonal constraint according to claim 1, wherein in the step S2), the overcomplete dictionary comprises a Laplacian wavelet dictionary, a Mo Laixiao wave dictionary and a harmonic wavelet dictionary.
5. The rolling bearing diagnosis method based on orthogonal projection mechanism and time domain orthogonal constraint of claim 1, wherein in step S3), an orthogonal dictionary matching pursuit ODMP algorithm based on the orthogonal projection mechanism is adopted, and atomic orthogonality is utilized to calculate a sparse coefficient through single inner product projection, and specifically comprises the following steps: S311), input vibration signal Orthogonal dictionary satisfying atomic orthogonality Presetting a frequency range and initializing a coefficient set Setting a stopping criterion; s312), calculate vibration signal Orthogonal dictionary Relativity of atoms in ; S313), if the stopping criterion is a fixed iteration number Selecting the correlation with the largest amplitude Index set corresponding to each atom Directly determining sparse coefficients ; S314), if the stopping criterion is the target residual error, calculating the coefficient of the selected atom And calculate the residual signal ; S315), judging whether the L2 norm of the residual signal is smaller than the threshold I.e. judging If yes, stopping iteration, otherwise repeating the steps S23-S24); S316), output sparse coefficient Residual signal For signal reconstruction.
6. The method for diagnosing the rolling bearing based on the orthogonal projection mechanism and the time domain orthogonal constraint according to claim 1, wherein in the step S3), a time domain orthogonal matching pursuit TOMP algorithm based on the time domain orthogonal constraint is adopted, and the selected atoms are forced to be orthogonal by utilizing the time non-overlapping property of fault impact, and the method specifically comprises the following steps: S321), input vibration signal Overcomplete dictionary Compact atomic support length Initializing an index set Initial residual signal Number of iterations ; S322), determining the optimal parameters of the overcomplete dictionary by the correlation filtering CF method, namely: constructing candidate original subset in the range of preset parameters, and calculating vibration signals Selecting parameters corresponding to candidate atoms with the largest correlation coefficient as optimal parameters according to the normalized correlation coefficients of the candidate atoms; S323), calculating the current residual Overcomplete dictionary Correlation vector of all atoms in (3) ; S324), when the stop condition is not satisfied, screening the correlation vector Atomic index with maximum medium amplitude Will index Adding index sets Updating ; S325), eliminating intra-neighborhood correlation, atoms The relevance vector element in the neighborhood of (1) is set to 0, forcing the selected atom to remain orthogonal to the neighborhood atom, i.e.: Wherein, the Representing a compact support length; s326), judging whether the stopping criterion is met, stopping iteration if the stopping criterion is met, and outputting the sparse coefficient with the support set of I 。
7. The method for diagnosing a rolling bearing based on an orthogonal projection scheme and time domain orthogonal constraint according to claim 6, wherein in step S322), said parameters include atomic frequency parameters And damping ratio In the preset parameter range, the atomic frequency parameters Meets the Nyquist sampling criterion, and the damping ratio The range is limited to [0, 0.2], the atomic compact support length And determining according to the minimum time interval of fault impact.
8. The method for diagnosing a rolling bearing based on an orthogonal projection mechanism and time domain orthogonal constraint of claim 7, wherein the time non-overlapping nature of the fault impact is such that the time interval of the fault impact events of the rolling bearing is greater than the atomic compact support length Ensuring that the support sets of any two selected atoms do not intersect each other.
9. The rolling bearing diagnosis method based on orthogonal projection mechanism and time domain orthogonal constraint of claim 8, wherein in step S4), the expression of reconstructing the vibration signal by using the sparse coefficient of the orthogonal dictionary matching pursuit ODMP algorithm is: ; In the formula, Is the first of orthogonal dictionary An atom; the expression of the sparse coefficient reconstruction vibration signal by using the time domain orthogonal matching pursuit TOMP algorithm is as follows: In the formula, For the reconstructed vibration signal(s), Representing an overcomplete dictionary set; representing a sparse coefficient set; Represent the first First iteration Sparse coefficients of individual atoms; representing the first of the complete dictionary And atoms.
10. A rolling bearing diagnostic system based on an orthogonal projection mechanism and a temporal orthogonal constraint, comprising: the signal acquisition module acquires vibration signals of the rolling bearing through a sensor, wherein the vibration signals comprise fault impact components and background noise; The dictionary adaptation construction module is used for analyzing noise intensity and frequency distribution characteristics of the signals based on the preprocessed signals output by the signal acquisition module, adaptively selecting an orthogonal dictionary or an overcomplete dictionary, and completing dictionary parameter optimization and dictionary construction according to preset rules; The sparse coefficient solving module is internally provided with an orthogonal dictionary matching pursuit ODMP algorithm and a time domain orthogonal matching pursuit TOMP algorithm, and according to the dictionary type output by the dictionary adaptation construction module, the corresponding algorithm is automatically invoked to solve the sparse coefficient; the signal reconstruction module is used for completing vibration signal reconstruction based on the sparse coefficient output by the sparse coefficient solving module and combining with the dictionary output by the dictionary adaptation construction module; The fault intelligent recognition module is used for carrying out envelope spectrum analysis on the reconstructed signal, extracting fault characteristic frequency FCF, harmonic component and modulation sideband, comparing with a preset rolling bearing fault frequency database and outputting fault type, fault characteristic parameter and diagnosis confidence.

Description

Rolling bearing diagnosis method and system based on orthogonal projection mechanism and time domain orthogonal constraint Technical Field The invention relates to the technical field of rolling bearing fault diagnosis, in particular to a rolling bearing diagnosis method and system based on orthogonal projection mechanism and time domain orthogonal constraint. Background The rolling bearing is used as a core key component of rotary machinery, is widely applied to the fields of machine tools, motors, rail transit, aerospace and the like, and the running state of the rolling bearing directly determines the reliability, safety and service life of equipment. In the whole life cycle of the rolling bearing, early faults are easily covered by background noise and harmonic interference generated by equipment operation due to weak impact energy and are difficult to directly identify, and if the faults are not diagnosed and processed in time, the early faults can be rapidly developed into serious faults, so that equipment is stopped, production is interrupted, and even safety accidents are caused. Therefore, the rapid and accurate diagnosis of the early failure of the rolling bearing is achieved. The sparse representation technology becomes one of the main stream technologies for reconstructing the fault impact components of the rolling bearing by virtue of the strong noise suppression and feature extraction capability, and the orthogonal matching pursuit OMP algorithm is widely used for solving the sparse coefficient due to flexible realization and excellent sparse recovery performance. However, conventional OMP algorithms and existing modification variants have the following core drawbacks, limiting their engineering applications: 1) the traditional OMP algorithm needs to execute pseudo-inverse calculation (such as SVD decomposition) and repeated inner product projection operation each iteration, the calculation complexity grows twice along with the signal length L and the dictionary size N, long-time vibration signals or large-scale data are difficult to process, 2) the dictionary and signal characteristics are not fully utilized, namely, for an orthogonal dictionary (such as DCT and Fourier basis), the OMP algorithm does not utilize atomic orthogonality to cause redundant calculation, for an overcomplete dictionary (such as a wavelet dictionary), the physical characteristics of sparse and non-overlapping fault impact time domains are not combined, complex iteration solution is still relied, 3) the real-time performance is insufficient, the traditional OMP algorithm and improved variants are difficult to deploy in an industrial real-time monitoring scene due to high calculation cost, and are especially not suitable for online diagnosis of high-speed bearings (such as train wheel pair bearings and aeroengine bearings), 4) the prior knowledge or manual intervention is relied on, a part of the improved schemes need to construct a fault model in advance, experience parameters are set, or learning training processes are relied on, and the self-adaption capability is poor, and the traditional OMP algorithm is not suitable for complex and variable working conditions. Therefore, development of a rolling bearing diagnosis method and system capable of fully utilizing dictionary structural characteristics and signal physical characteristics, remarkably reducing calculation complexity on the premise of maintaining diagnosis accuracy, and having strong self-adaptive capability and engineering practicability is needed. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a rolling bearing diagnosis method and a system based on orthogonal projection mechanism and time domain orthogonal constraint, the invention designs an optimization algorithm aiming at an orthogonal dictionary and an overcomplete dictionary respectively, by introducing an orthogonal projection mechanism and time domain orthogonal constraint, redundant calculation of the traditional OMP algorithm is fundamentally eliminated, the calculation efficiency is greatly improved while the diagnosis accuracy is ensured, and the rapid and accurate diagnosis of the rolling bearing faults is realized. In order to achieve the above purpose, the present invention adopts the following technical scheme: In a first aspect, the present invention provides a rolling bearing diagnosis method based on orthogonal projection mechanism and time domain orthogonal constraint, comprising the steps of: s1), collecting vibration signals of a rolling bearing through a sensor, wherein the vibration signals comprise fault impact components and background noise; s2), adaptively selecting an orthogonal dictionary or an overcomplete dictionary according to the intensity and frequency distribution characteristics of the vibration signals, and completing dictionary parameter optimization and dictionary construction according to preset rules; s3), calculating sparse