CN115630280A

CN115630280A - Rolling bearing fault diagnosis method based on CEEMD multi-scale diffusion entropy and PSO-ELM

Info

Publication number: CN115630280A
Application number: CN202211407795.9A
Authority: CN
Inventors: 毛美姣; 向林; 王建涛; 肖文强; 杜光超
Original assignee: Xiangtan University
Current assignee: Xiangtan University
Priority date: 2022-11-10
Filing date: 2022-11-10
Publication date: 2023-01-20

Abstract

The invention discloses a rolling bearing fault diagnosis method based on combination of complementary set empirical mode decomposition (CEEMD) multi-scale dispersion entropy and Particle Swarm Optimization (PSO) optimized Extreme Learning Machine (ELM). The method comprises the steps of firstly decomposing a non-stationary original acceleration vibration signal of the rolling bearing by adopting a CEEMD method to obtain a plurality of stationary Intrinsic Mode Function (IMF) components, screening the IMF components according to a correlation coefficient principle, then extracting the characteristics of the screened IMF components by utilizing multi-scale dispersion entropy, finally sending the characteristic set into an ELM optimized by PSO for classification and identification, and finally realizing the extraction and diagnosis of fault characteristics. The method has the advantages of achieving a good effect on the Kaiser university data set, being high in operation speed and high in anti-interference capability, solving the problem of inaccurate classification caused by random selection of traditional ELM parameters, and greatly improving the fault diagnosis precision.

Description

Rolling bearing fault diagnosis method based on CEEMD multi-scale diffusion entropy and PSO-ELM

Technical Field

The invention belongs to the field of mechanical fault diagnosis, and relates to a fault diagnosis method of a rolling bearing.

Background

Rolling bearings are one of the essential components in many mechanical equipment, playing an important role, and at the same time, they are one of the weakest ones. Once damaged, the bearing has great influence on the service life and performance of the machine, and even has great accidents, which causes great economic loss and endangers personal safety. Researches show that the rolling bearing fault accounts for 45-55% of the total fault of mechanical equipment and seriously influences the operation efficiency of the machinery, so that the research on the early fault diagnosis method of the rolling bearing has great significance for improving the safety of the operation of the machinery and predicting the fault avoidance risk in advance. Due to the fact that vibration signals of the rolling bearing are non-linear and non-stationary, a large number of typical fault samples are difficult to obtain, and therefore an effective fault diagnosis method is a hot spot of current research.

In order to make the fault diagnosis result more accurate, YEH J proposes a novel signal processing method, namely a complementary set empirical mode decomposition (CEEMD) method, which not only can effectively suppress the mode confusion problem generated by the Empirical Mode Decomposition (EMD) method, but also has shorter operation time than the integrated empirical mode decomposition (EEMD), and thus is widely applied to mechanical fault diagnosis. The signal after CEEMD decomposition needs feature extraction, and under the efforts of researchers, a plurality of feature extraction methods appear, for example: the method comprises the steps of sample entropy, permutation entropy, fuzzy entropy, symbol entropy, basic scale entropy, dispersion entropy and the like, and the methods can measure the complexity of signals and effectively improve the fault diagnosis precision. However, the entropy is usually a measurement performed on a signal in a single scale, and the fault information in the fault signal of the rolling bearing is usually distributed in different scales, and the measurement performed only in the single scale is likely to cause the loss of the fault information, so that the fault information cannot be accurately depicted. Therefore, the feature extraction method which can overcome the defect of fault information loss and has strong anti-interference capability is the key for improving the fault diagnosis precision.

With the advent of the digital era, the machine learning method can solve the problems of prediction, classification, clustering, feature extraction and the like, wherein the effect of solving the classification problem is remarkable, and an Extreme Learning Machine (ELM) is a classification method proposed by Huang and the like. However, since the initial value and the threshold of the ELM algorithm are random, uncertainty is brought to the model, and the classification accuracy is affected. Therefore, optimizing the method for selecting parameters by the ELM to improve the classification accuracy is an urgent problem to be solved.

Disclosure of Invention

In order to overcome the defects, the invention provides a rolling bearing fault diagnosis method based on the combination of CEEMD multi-scale dispersion entropy and PSO-ELM.

In order to achieve the purpose, the invention adopts the following technology to realize the purpose:

a rolling bearing fault diagnosis method based on combination of CEEMD multi-scale diffusion entropy and PSO-ELM. The method comprises the following steps:

s1: and acquiring an original signal of the rolling bearing.

S2: the original signal is decomposed using CEEMD to obtain a plurality of IMF components.

S3: and screening the IMF components by adopting a correlation coefficient principle.

S4: and (4) performing feature extraction on the components after S3 screening by adopting a multi-scale dispersion entropy.

S5: and optimizing ELM network parameters by adopting a PSO optimization algorithm.

S6: and (4) training the rolling bearing feature set obtained in the front by using the optimized ELM to realize accurate classification of fault features.

In step S2, complementary Ensemble Empirical Mode Decomposition (CEEMD) is performed on the acquired original vibration signal to obtain a plurality of Intrinsic Mode Function (IMF) components.

In step S3, because the eigen-mode function component of S2 has a certain correlation with the original signal, the correlation coefficient between each eigen-mode function component and the original signal is calculated by using the correlation coefficient principle, so as to screen out the eigen-mode function component with the largest correlation, and then recombine the eigen-mode function component into a new vibration signal, which removes noise interference and makes the vibration signal purer compared with the original signal.

In step S4, the improved scatter entropy (DE), i.e. the multi-scale scatter entropy, is used to perform fault feature extraction, so as to form a feature set, and the multi-scale scatter entropy has a higher advantage than the scatter entropy in measuring time series complexity and irregularity, and is very suitable for performing signal feature extraction.

In step S5, since the initial value and the threshold of the Extreme Learning Machine (ELM) algorithm are random, which may affect the training result to a certain extent, the Particle Swarm Optimization (PSO) algorithm is used to search out the optimal parameters, and the specific operation is to set the ELM input weight IW and the hidden layer neuron bias IB as the particles of the PSO algorithm, so as to avoid the random training of the ELM model.

In step S6, the improved extreme learning machine is used to train the feature set, and finally, accurate classification of the fault is obtained.

In the optimized ELM network parameters described in S5, the input weight IW and the hidden layer neuron bias IB are set as particles of the PSO algorithm, so that random training of an ELM model is avoided. The implementation process comprises the following steps:

first, initializing population size N ₁ And the number of times of population update it max.

And secondly, randomly generating an inertia parameter w according to sample data, taking the mean square error of the ELM test sample output and the predicted output as a fitness function, calculating the fitness value of each particle, updating the position and the speed of the particle through comparison and optimization, and finally obtaining the ELM network parameter after particle swarm optimization when the mean square error is minimum or reaches the maximum iteration times.

The invention has the characteristics and beneficial effects that:

1. the invention adopts CEEMD method, which can not only effectively inhibit the problem of mode confusion, but also shorten the operation time.

2. The feature extraction is carried out by adopting the multi-scale diffusion entropy, the operation speed is high, the anti-interference capability is strong, the defect that fault information is easy to lose due to single scale is overcome, and the precision of fault diagnosis is effectively improved.

3. And the ELM network parameters are optimized by adopting a PSO algorithm, random training of an ELM model is avoided, and the classification precision is improved.

4. The fault diagnosis method is combined with the optimized fault classification method, so that the fault diagnosis efficiency is effectively improved.

Drawings

FIG. 1 is a flow chart of the present invention;

FIG. 2 is a CEEMD decomposition flow diagram;

FIG. 3 is a graph of IMF components of one set of CEEMD decomposed samples;

FIG. 4 is a graph of correlation coefficients of IMF components with respect to an original signal;

FIG. 5 is a PSO-ELM iteration chart;

FIG. 6 is an ELM confusion matrix before and after PSO optimization;

FIG. 7 is a PSO-ELM confusion matrix under different loads.

Detailed Description

The invention relates to a rolling bearing fault diagnosis method based on combination of CEEMD multi-scale diffusion entropy and PSO-ELM, a flow chart of which is shown in figure 2, and the process comprises the following steps:

the CEEMD decomposition process comprises the following steps:

step 1, introducing a group of additive and subtractive white noise signals into an original signal x (t) to obtain:

wherein H _m 、J _m Is the signal sequence after white noise is added. n is _m (t) represents white noise added at the mth time.

Step 2 Using EMD vs H _m 、J _m And decomposing to obtain IMF components of the two, specifically as follows:

c _j，m and (t) represents the jth IMF component obtained after the mth white noise addition and EMD decomposition, and q is the number of IMFs.

Step 3, repeating the step 1 and the step 2, and calculating the average value of IMF components obtained after m times of decomposition, namely the jth IMF component c _j

(2) The multi-scale dispersion entropy feature extraction method comprises the following steps:

step 1, the time sequence of the original signal is { u (i), i =1, 2.. Multidata., L }, the sequence is coarsely granulated compositely, and the k-th coarse-grained sequence under the scale factor tau is coarsely granulated by x _k ^τ Expressed, the following is an expression of the sequence:

step 2, calculating DE (x) of each coarse graining sequence according to the DE principle under each scale factor tau respectively _k ^τ M, c, d), then MDE is expressed as:

where c is the category, d is the time delay, m is the embedding dimension, and X is the initial time series signal.

(3) Particle swarm optimization ELM network parameter step:

step 1, initializing population size N ₁ And the number of times of population update it max.

And 2, randomly generating an inertia parameter w according to sample data, taking the mean square error of the ELM test sample output and the predicted output as a fitness function, calculating the fitness value of each particle, updating the position and the speed of the particle through comparison and optimization, and finally obtaining the ELM network parameter after particle swarm optimization when the mean square error is minimum or reaches the maximum iteration times.

In order to verify the fault diagnosis rate of the fault diagnosis combination method, the rolling bearing data set of the Kaiser university is adopted, matlab is used for testing, a driving end bearing is selected as an experimental bearing, the fault diameter is 0.007 inches, the sampling frequency is 12kHz, the motor rotating speed is 1797rpm, the experiment is divided into four groups of 0, 1,2 and 3 according to the motor load, 10 types of experiments are performed in each group, each type comprises 40 samples, and the vibration signal data description is shown in table 1.

TABLE 1 vibration signal data

Firstly, CEEMD decomposition with short operation time and capability of effectively inhibiting modal confusion is carried out, wherein a component diagram is shown in figure 3; each component has certain correlation with the original signal, the magnitude of the correlation is measured through a correlation coefficient, the correlation coefficient of each IMF component and the original signal is obtained and is shown in figure 4, and a multi-scale dispersion entropy is used for extracting characteristic values to form a characteristic matrix; an iteration number graph of the PSO optimization extreme learning machine is drawn as shown in FIG. 5, and it can be seen that convergence is achieved when the PSO algorithm iterates to 73 rd time, the optimal value of the fitness function is 0.001, and the optimal parameters obtained at this time are substituted into the ELM. And then, performing state recognition on the feature set by using PSO-ELM, comparing the feature set with an unoptimized extreme learning machine, as shown in fig. 6, greatly reducing the number of wrongly recognized samples after optimization, obviously increasing the accuracy, finally performing test tests under four loads of a Kaesi university data set, as shown in fig. 7, and obviously showing that the algorithm accuracy reaches very high accuracy under four conditions as shown in table 2, so that the rolling bearing fault diagnosis method based on combination of CEEMD multi-scale entropy distribution and PSO-ELM can improve the fault diagnosis efficiency.

TABLE 2 PSO-ELM accuracy under different loads

Claims

1. A rolling bearing fault diagnosis method based on CEEMD multi-scale diffusion entropy and PSO-ELM is characterized by comprising the following steps:

s1: acquiring an original signal of a rolling bearing;

s2: decomposing an original signal by adopting CEEMD to obtain a plurality of IMF components;

s3: screening IMF components by adopting a correlation coefficient principle;

s4: carrying out feature extraction on the components screened in the S3 by adopting a multi-scale diffusion entropy;

s5: optimizing ELM network parameters by adopting a PSO optimization algorithm;

2. The CEEMD multi-scale entropy and PSO-ELM-based rolling bearing fault diagnosis method of claim 1, wherein in step S2, a complementary set empirical mode decomposition (CEEMD) is performed on the acquired original vibration signal to obtain a plurality of Intrinsic Mode Function (IMF) components.

3. The method as claimed in claim 1, wherein in step S3, since the eigenmode function component of S2 has a certain correlation with the original signal, the correlation coefficient between each eigenmode function component and the original signal is calculated by using the correlation coefficient principle, so as to screen out the eigenmode function component with the largest correlation, and then the eigenmode function component is recombined into a new vibration signal, and the new vibration signal is compared with the original signal, so as to remove noise interference and make the vibration signal more pure.

4. The CEEMD multi-scale entropy diffusion and PSO-ELM based rolling bearing fault diagnosis method as claimed in claim 1, wherein in step S4, improved entropy Diffusion (DE) is adopted, namely multi-scale entropy diffusion is adopted to perform fault feature extraction, a feature set is formed, and the multi-scale entropy diffusion is more prominent than the entropy diffusion in measuring time series complexity and irregularity, and is very suitable for signal feature extraction.

5. The CEEMD multi-scale entropy and PSO-ELM based rolling bearing fault diagnosis method of claim 1, wherein in step S5, since both the initial value and the threshold of the Extreme Learning Machine (ELM) algorithm are random, which has a certain effect on the training result, the Particle Swarm Optimization (PSO) algorithm is used to search out the optimal parameters, and the specific operation is to set the ELM input weight IW and the hidden layer neuron bias IB as the particles of the PSO algorithm, so as to avoid the random training of the ELM model.

6. The CEEMD multi-scale entropy and PSO-ELM based rolling bearing fault diagnosis method of claim 1, wherein in step S6, the feature set is trained by using an improved extreme learning machine, and finally, an accurate classification of the fault is obtained.