CN-121994486-A - Rolling bearing fault diagnosis method based on parameter optimization FMD and self-adaptive denoising

CN121994486ACN 121994486 ACN121994486 ACN 121994486ACN-121994486-A

Abstract

The invention discloses a rolling bearing fault diagnosis method based on parameter optimization FMD and self-adaptive denoising, which comprises the steps of acquiring bearing vibration signals in different running states based on an acceleration sensor, searching and optimizing FMD decomposition parameters by adopting an improved long-nose raccoon optimization algorithm, carrying out FMD decomposition on the vibration signals based on optimal parameter combination to obtain a plurality of IMF components, analyzing each IMF component by a Teager energy operator, dividing each IMF component into a noise dominant component and a signal dominant component according to a discriminant, carrying out self-adaptive denoising treatment on the noise dominant component, reconstructing the noise dominant component and the signal dominant component to obtain a reconstructed signal, calculating a Teager energy operator on the reconstructed signal, constructing a Teager energy spectrum, realizing extraction of rolling bearing fault characteristics and identification of fault types, effectively reducing subjectivity of parameter selection and influence of noise interference on diagnosis results, and improving stability and accuracy of rolling bearing fault diagnosis.

Inventors

HUA XIAOPENG
WANG TAOTAO
XU SEN

Assignees

盐城工学院

Dates

Publication Date: 20260508
Application Date: 20260126

Claims (10)

1. A rolling bearing fault diagnosis method based on parameter optimization FMD and self-adaptive denoising is characterized by comprising the following steps: acquiring bearing vibration signals of different running states based on an acceleration sensor; The improved long-nose raccoon optimization algorithm is used for constructing an elite set, calculating refraction candidate solutions of elite individuals, updating based on greedy criteria, generating refraction candidate solutions through Gaussian variation processing of the individuals in the elite set, updating based on greedy criteria, and realizing parameter optimization by combining population guidance mode control, boundary restoration and fitness sequencing screening, wherein the decomposition parameters of the FMD are used for controlling the length L of a filter, the frequency band division number C and the modal number M in the FMD decomposition process; based on the optimal parameter combination, carrying out FMD decomposition on the bearing vibration signal to obtain a plurality of intrinsic mode function IMF components; analyzing and classifying the IMF component according to a discrimination criterion based on Teager energy median, and dividing the IMF component into a noise dominant component and a signal dominant component; performing self-adaptive denoising treatment on the noise dominant component to obtain a denoised IMF component; reconstructing the denoised IMF component and the signal dominant component to obtain a reconstructed signal; And performing Teager energy operator calculation on the reconstruction signal, and drawing a Teager energy spectrum to realize extraction of fault characteristics of the rolling bearing and identification of fault types.
2. A method for diagnosing a rolling bearing fault based on a parameter optimized FMD and adaptive denoising as claimed in claim 1, wherein the bearing vibration signal comprises: ; Wherein, the For the moment of time Is used for measuring the sampling value of the vibration of the bearing, For the number of sample points, For the sampling frequency to be the same, Is the sampling time and satisfies 。
3. The rolling bearing fault diagnosis method based on parameter optimization FMD and adaptive denoising as claimed in claim 1, wherein searching and optimizing the decomposition parameters of FMD based on improved long-nose raccoon optimization algorithm, determining the optimal parameter combination of the bearing vibration signal comprises: Initializing population members, wherein each member corresponds to three parameters of filter length, frequency band division number and modal number; Calculating the fitness value of each population member, constructing a fitness sequence of an initial population, determining a global optimal individual based on the fitness sequence of the initial population, and initializing a convergence record, wherein the fitness value is envelope spectrum entropy ESE of the biggest kurtosis component in M modal components after FMD decomposition; Determining a guiding mode according to the index of each population member, wherein the guiding mode comprises global optimal position guiding and random guiding; performing boundary restoration on candidate positions of each population member, and determining search boundary constraint; sorting the fitness values corresponding to the population members based on the small-to-large order, and determining a first sorting index sequence; selecting a first elite set formed by a plurality of individuals in a first ordering index sequence; Calculating a corresponding refraction candidate solution for each elite individual in the first elite set, calculating an adaptability value corresponding to the refraction candidate solution, and updating based on a greedy criterion to obtain an updated ordering index sequence as a second ordering index sequence; The method comprises the steps of obtaining a preset elite set scale coefficient, determining a second elite set rule number based on the preset elite set scale coefficient, selecting individuals of a previous second elite set rule number in a second sequencing index sequence to form a second elite set; performing Gaussian variation on each elite individual in the second elite set, generating each elite individual, calculating a corresponding refraction candidate solution, calculating an adaptability value corresponding to the refraction candidate solution, and updating based on a greedy criterion to obtain an updated second ordering index sequence as a third ordering index sequence; calculating an optimal fitness value in the third ordered index sequence; And comparing the optimal fitness value with a global optimal fitness value, and determining an optimal parameter combination of the bearing vibration signal when the optimal fitness value is determined to be smaller than the global optimal fitness value.
4. The rolling bearing fault diagnosis method based on parameter optimization FMD and adaptive denoising according to claim 1, wherein performing FMD decomposition on the bearing vibration signal based on the optimal parameter combination to obtain a plurality of eigenmode function IMF components comprises: Dividing a frequency band into C sub-frequency bands in a normalized frequency interval [0,1 ] according to optimal decomposition parameters L, C and M in the optimal parameter combination, wherein L is the length of a filter, C is the number of initial sub-frequency bands, and M is the number of target modes; sequentially executing outer layer iteration rounds on each initial channel signal in the bearing vibration signals; After each round of outer layer iteration is finished, a channel correlation matrix is constructed based on the output signals of all the current initial channels, wherein the off-diagonal element is the absolute value of the correlation coefficient of any two channels, and the diagonal element is set to zero; Judging whether the number of the residual channels is equal to M-1, if so, terminating iteration, and taking the output signals of the residual channels as intrinsic mode function IMF components.
5. The rolling bearing fault diagnosis method based on parameter optimization FMD and adaptive denoising according to claim 1, wherein analyzing and classifying the IMF component according to discrimination criteria based on Teager energy median, dividing the IMF component into noise dominant component and signal dominant component, comprises: calculating a discrete Teager energy operator sequence corresponding to each IMF component; Calculating a Teager energy accumulation value of each modal component based on the discrete Teager energy operator sequence; constructing a Teager energy accumulation value set based on Teager energy accumulation values of all modal components; Constructing a binary discrimination mask based on the adaptive discrimination threshold; the IMF component is divided into noise dominant components and signal dominant components based on the binary discrimination mask.
6. The rolling bearing fault diagnosis method based on parameter optimization FMD and adaptive denoising as claimed in claim 1, wherein performing adaptive denoising processing on the noise dominant component to obtain a denoised IMF component comprises: taking a discrete sequence corresponding to the noise dominant component as input to construct an initialization set; setting multiple coefficients, carrying out iterative rounds on the initialization set, and calculating the mean value and standard deviation corresponding to the initialization set; determining a noise threshold based on the multiple coefficient and standard deviation; Iteratively screening the set elements based on the noise threshold value to determine a target threshold value; and denoising the noise dominant component based on the target threshold value to obtain a denoised IMF component.
7. The rolling bearing fault diagnosis method based on parameter optimization FMD and adaptive denoising as claimed in claim 1, wherein reconstructing the denoised IMF component and the signal dominant component to obtain a reconstructed signal comprises: ; Wherein, the Reconstructing the signal; is the dominant component of the signal; is the noise dominant component after denoising.
8. The rolling bearing fault diagnosis method based on parameter optimization FMD and adaptive denoising according to claim 1, wherein performing Teager energy operator calculation on the reconstructed signal and drawing a Teager energy spectrum comprises: Calculating a discrete Teager-Kaiser energy operator sequence corresponding to the reconstruction signal; performing mean value removal processing on the discrete Teager-Kaiser energy operator sequence to obtain a zero-mean value sequence; Performing discrete Fourier transform on the zero-mean sequence to obtain a frequency domain complex sequence; determining an energy spectral density index value based on the sequence of frequency domain complex numbers; Acquiring a unilateral Teager energy spectrum and determining a corresponding frequency coordinate; And drawing a Teager energy spectrum curve by taking the frequency coordinate as a horizontal axis and the energy spectrum value as a vertical axis.
9. The rolling bearing fault diagnosis method based on parameter optimization FMD and adaptive denoising according to claim 5, wherein calculating the discrete Teager energy operator sequence corresponding to each IMF component comprises: ; Wherein, the Represent the first The modal components are at the sampling point The Teager energy operator output value; Represent the first The modal components are at the sampling point A value at; sampling point sequence number; Is an index of modal components.
10. A method for diagnosing a rolling bearing fault based on a parameter optimized FMD and adaptive denoising as claimed in claim 3, wherein calculating a corresponding refraction candidate solution for each elite individual in the first elite set comprises: ; Wherein, the A j-th dimensional parameter solution which is a refraction candidate solution; Searching the midpoint of the interval for the j-th dimension parameter; Is the refractive index; index of elite individuals in the first elite set; Is the first elite set The j-th dimensional parameter value of each elite individual.

Description

Rolling bearing fault diagnosis method based on parameter optimization FMD and self-adaptive denoising Technical Field The invention relates to the technical field of mechanical equipment state monitoring and fault diagnosis, in particular to a fault diagnosis method of a rolling bearing based on parameter optimization FMD and self-adaptive denoising. Background Rolling bearings are used as key components in rotary machines in a wide range of applications, and their operating state directly affects the safety and reliability of the mechanical system. In actual working conditions, the rolling bearing is easily affected by factors such as load change, worsening of lubrication condition, environmental noise and the like, so that different types of faults are generated at the positions of the inner ring, the outer ring or the rolling body. Therefore, how to accurately extract the fault characteristics of the rolling bearing from the vibration signals under the complex and strong noise background is always the research focus in the field of mechanical fault diagnosis. The existing rolling bearing fault diagnosis method is mostly based on vibration signal analysis, and the state of the rolling bearing is evaluated through means such as time domain analysis, frequency domain analysis and time frequency analysis. Among them, the modal decomposition type method is widely focused on being able to decompose a non-stationary and nonlinear signal into several modal components having physical significance. However, the conventional modal decomposition method generally has the problems of dependence of decomposition parameters on experience, insufficient stability of decomposition results, modal aliasing and the like in practical application, and effective fault characteristics are difficult to stably extract under complex working conditions. The characteristic modal decomposition (Feature Mode Decomposition, FMD) is used as a novel signal decomposition method, and the modal aliasing problem of the traditional method is improved to a certain extent by constructing a filter bank and iteratively extracting narrow-band modal components. However, in practical application, the FMD decomposition effect still highly depends on the selection of decomposition parameters, the decomposition results under different parameter combinations have large differences, and incorrect parameter selection easily results in weakening or covering effective fault information, so that an optimization algorithm is urgently needed to determine the selection of parameters. In order to improve the adaptability of the modal decomposition method under the complex working condition, in recent years, research attempts are made to introduce an intelligent optimization algorithm to automatically optimize decomposition parameters. The group intelligent optimization algorithm is widely applied to the fields of signal processing and mechanical fault diagnosis because of the characteristics of no gradient information, suitability for nonlinear and multimodal optimization problems and the like. For example, by introducing a swarm intelligent optimization algorithm to optimize modal decomposition parameters, the decomposition effect can be improved to a certain extent and the extractability of fault features can be enhanced. The long-nose raccoon optimization algorithm (Coati Optimization Algorithm, COA) has been proposed and applied in recent years to continuous optimization problems as a group intelligent optimization algorithm based on group collaboration and search behavior. The algorithm realizes global search of the target space by simulating the cooperative behavior of the long-nasal raccoon group in the search process, and has the characteristics of simple structure, fewer parameters and the like. Therefore, long-nose raccoon optimization algorithms and related variants thereof are increasingly being introduced into the application scenarios of parameter optimization and signal processing. However, in practical application, the COA as a novel intelligent optimization algorithm may still be affected by factors such as initial condition setting, single search strategy, insufficient population diversity, etc., and there are problems of unstable convergence speed, easy sinking into local optimization, etc. When the method is directly applied to FMD decomposition parameter optimization, the decomposition result can have larger fluctuation under different signals or working conditions, so that the stability and accuracy of the subsequent fault feature extraction and diagnosis result are affected. Therefore, how to combine the parameter characteristics of FMD to reasonably design and improve the optimization algorithm so as to improve the global searching capability and stability of the parameter optimizing process is still a problem to be solved in the prior art. In addition, rolling bearing vibration signals often contain a significant amount of ambient noise a