A Gear Vibration Fault Diagnosis Method

Abstract

The present invention provides a gear vibration fault diagnosis method. Considering the nonlinearity, non-stationarity and complexity of the operating environment of the gear vibration signal, a fault identification method combining multi-scale weighted permutation entropy MWPE and dung beetle algorithm DBO optimized support vector machine SVM is proposed from the two aspects of feature extraction and pattern recognition. The coefficient of variation is introduced to determine the parameters, and the feasibility and advantage of using the MWPE algorithm for gear fault feature extraction after the parameter optimization selection is verified. Then, the appropriate feature dimension is selected to construct a fault feature set, and the training set and the test set are further divided. The divided training set is brought into the DBO-SVM model to train the fault diagnosis model, and the trained model is used for prediction and classification of the test set to achieve the recognition of different gear fault patterns. The method has high accuracy, overcomes the problem that the feature extraction algorithm is unstable and easily falls into the local optimum, and ensures the accuracy of gear vibration fault diagnosis.

Description

Technical Field

[0001] The invention relates to the technical field of gear fault diagnosis, and in particular to a gear vibration fault diagnosis method. Background Art

[0002] Gears are a common part and an indispensable component of mechanical equipment, and are widely used in modern industry. Gear failure is one of the main causes of mechanical equipment failure, and timely detection and diagnosis are very important.

[0003] The characteristic carrier of gear fault is vibration signal, which is usually diagnosed by analyzing the vibration signal collected by the sensor. Since gears work under the joint action of multiple excitation sources and the working environment is complex, the obtained signal has the characteristics of nonlinearity and non-stationarity. How to extract fault characteristics and separate fault modes is a problem that needs to be solved.

[0004] In view of the problem of gear fault diagnosis, scholars at home and abroad have proposed many fault diagnosis methods. From the early time domain and frequency domain analysis methods, to the later time-frequency analysis method, and then to the current machine learning method. Support vector machines are widely used in the field of fault diagnosis because they have the advantages of high recognition accuracy and less time for small sample and low-dimensional data classification. However, their performance is easily affected by the penalty factor and kernel function parameters. Therefore, some researchers use simulated annealing algorithm, artificial fish swarm algorithm and particle swarm algorithm to optimize SVM parameters, but these methods take a long time to optimize and are prone to fall into local optimal solutions. There are certain defects, which limit the application and development of fault diagnosis technology. Summary of the invention

[0005] In view of this, an object of the present invention is to provide a gear vibration fault diagnosis method to at least solve the above problems.

[0006] The technical solution adopted by the present invention is as follows:

[0007] A gear vibration fault diagnosis method, the method comprising the following steps:

[0008] Step 1: Use a vibration acceleration sensor to collect gear vibration acceleration data of healthy gears and gears with different faults;

[0009] Step 2: Analyze the collected data and use the coefficient of variation to determine the parameters in the MWPE algorithm;

[0010] Step 3: Calculate the MWPE value of the data, classify and label the data with the calculated MWPE value according to the fault type, and divide the data of each fault type into training samples and test samples;

[0011] Step 4: Use the training samples to train the DBO-SVM classifier model;

[0012] Step 5: Use the trained DBO-SVM classifier model to identify the fault type of the test sample.

[0013] Furthermore, the faulty gears in step 1 include: missing tooth gears, cracked gears, peeled gears, and sharpened gears.

[0014] Furthermore, the parameters in the MWPE algorithm in step 2 include: sample length N, embedding dimension m, scale factor s, and delay time τ.

[0015] Furthermore, in step 2, the parameters in the MWPE algorithm are determined, specifically:

[0016] Set the scale factor s≥10, the sample length N should satisfy N≥1024, and the delay time τ=1;

[0017] A group of data is randomly selected from the gear vibration acceleration data of healthy gears and faulty gears, and the embedding dimension m=3, 4, 5, 6, 7 is taken. The parameters are respectively brought in and the MWPE value of each group of data is calculated. Then the standard deviation and average value of the MWPE values ​​of the two gear data are calculated. The coefficient of variation CV is calculated using the standard deviation and average value. The embedding dimension m when the coefficient of variation CV is the minimum value is selected as the final embedding dimension m value; the coefficient of variation CV calculation formula is:

[0018]

[0019] Among them, σ is the standard deviation and μ is the mean.

[0020] Furthermore, the MWPE value of the calculated data in step 3 is specifically:

[0021] First, calculate the WPE value of the data. The specific algorithm is as follows:

[0022] For a given signal sequence Multiple subsequences are reconstructed using phase space reconstruction theory. The formula is as follows:

[0023]

[0024] Where m is the embedding dimension and τ is the delay time;

[0025] Calculate each subsequence The weight value is as follows:

[0026]

[0027] Among them, ω i is the weight value, is the mean of each subsequence, x i+(t-1)τ is a single data point for each subsequence;

[0028] For each signal subsequence Use the weight value ω i and permutation and combination patterns Represents characteristic information; there are K kinds of signal sequence arrangement modes, and the probability weighted value of each mode is:

[0029]

[0030] in, for The arrangement mode N is the given signal length, is a subsequence of each signal;

[0031] Calculate the WPE value of the signal sequence, the formula is as follows:

[0032]

[0033] in, The probability weighted value for each mode, is a subsequence of each signal, m is the embedding dimension, and τ is the delay time;

[0034] Then, for the signal sequence x i The coarse-grained sequence is obtained by coarse-grained processing, and the formula is as follows:

[0035]

[0036] Where s is the scale factor and N is the given signal length;

[0037] Finally, the coarse-grained sequence is used to calculate the WPE value at multiple scales, that is, the MWPE value, and the formula is as follows:

[0038] MWPE(X,τ,s,m)=WPE(y s (j),m,τ)

[0039] Normalization is performed, and the formula is as follows:

[0040]

[0041] Furthermore, in step 3, the data of each fault type are divided into training samples and test samples, specifically: the data of each fault type are divided into training samples and test samples in a ratio of 4:6.

[0042] Furthermore, in step 4, the DBO-SVM classifier model is trained using training samples, specifically:

[0043] The training samples are input into the SVM model, and the fitness function is set as the minimum SVM recognition error rate. The optimal parameters of the SVM penalty factor c and kernel function parameter g are selected through the dung beetle algorithm, and the SVM model is trained using the selected optimal parameters.

[0044] Further, the specific steps are:

[0045] Step S1: input training samples, initialize SVM and DBO parameters, use mapminmax function to normalize sample data to between 0 and 1, use the minimum SVM model recognition error rate as the fitness function, and set the maximum number of DBO iterations to T;

[0046] Step S2: updating the position and fitness function according to the updating rules of different dung beetles. If the new position is better than the original position, the new position replaces the original position, otherwise the original position remains unchanged.

[0047] Step S3: Determine whether the maximum number of iterations is met. If not, return to step S2; if so, output the parameters of the penalty factor c and kernel function parameter g of the SVM, which are the optimal parameters.

[0048] Step S4: Use the optimal parameters of the SVM penalty factor c and the kernel function parameter g to train the SVM model.

[0049] Furthermore, step 5 is specifically as follows: inputting the test sample into the DBO-SVM classifier model, the DBO-SVM classifier model processes the data in the test sample and matches it with the fault type, and outputs the corresponding fault type.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] In view of the difficulty in extracting gear fault features, the present invention proposes a multi-scale weighted permutation entropy to characterize the gear fault features. Since it is difficult to determine the embedding dimension of the multi-scale weighted permutation entropy parameters, the coefficient of variation method is introduced to analyze and determine it; in the fault identification stage, since the performance of the support vector machine is easily affected by the penalty factor and the kernel function parameters, the dung beetle optimization algorithm considers both global exploration and local development, with fast convergence speed, good stability and high solution accuracy. It is applied to the optimization of support vector machine parameters, which is conducive to accurately identifying the fault type. This method has high recognition accuracy, small calculation amount and high recognition efficiency, ensuring the accuracy of gear fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only preferred embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0053] Figure 1 is a schematic diagram of a fault diagnosis method flow chart of an embodiment of the present invention;

[0054] Figure 2 It is a schematic diagram of the DBO-SVM classifier model training process of an embodiment of the present invention;

[0055] Figure 3 MWPE analysis result diagram of a healthy gear and a tooth-missing gear according to an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The technical solution of the present invention is further elaborated in detail below in conjunction with the drawings and specific embodiments of the specification. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention. In the following description, the expression "some embodiments" is related to a subset of all possible embodiments, but it should be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0057] In the following description, a large number of specific details are provided to provide a more thorough understanding of the present invention. However, it is apparent to those skilled in the art that the present invention can be implemented without one or more of these details. In other examples, in order to avoid confusion with the present invention, some technical features well known in the art are not described.

[0058] It should be understood that the present invention can be implemented in different forms and should not be interpreted as being limited to the embodiments proposed herein. On the contrary, providing these embodiments will make the disclosure thorough and complete, and the scope of the present invention will be fully conveyed to those skilled in the art. And the purpose of the terms used herein is only to describe specific embodiments and is not intended to be a limitation of the present invention. When used herein, the singular forms of "one", "one" and "said / the" are also intended to include plural forms, unless the context clearly indicates another way. It should also be understood that the terms "compose" and / or "include" when used in this specification determine the presence of the features, integers, steps, operations, elements and / or parts, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, parts and / or groups. When used herein, the term "and / or" includes any and all combinations of the relevant listed items.

[0059] It should also be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "inside", "outside", "left", "right" and similar expressions used herein are for illustrative purposes only and are not intended to be the only implementation method.

[0060] In order to fully understand the present invention, a detailed structure will be proposed in the following description to illustrate the technical solution proposed by the present invention. The optional embodiments of the present invention are described in detail as follows, but in addition to these detailed descriptions, the present invention may also have other implementations.

[0061] Reference Figure 1 The present invention provides a gear vibration fault diagnosis method, the method comprising the following steps:

[0062] Step 1: Use a vibration acceleration sensor to collect gear vibration acceleration data of healthy gears and gears with different faults;

[0063] The faulty gears include: missing tooth (MT), root crack (RC), spalling (SP), and chipping tip (CT); chipping tip gears have five types of damage degrees, which are represented by CT1, CT2, CT3, CT4, and CT5 respectively.

[0064] The specific information collected in this embodiment is shown in Table 1.

[0065]

[0066] Table 1 Gear data information

[0067] Step 2: Analyze the collected data and use the coefficient of variation to determine the parameters in the MWPE algorithm;

[0068] Exemplarily, the MWPE algorithm requires four key parameters to be set manually: sample length N, embedding dimension m, scale factor s, and delay time τ. Among them, there is currently no selection standard for the scale factor s, and it is generally set to be greater than 10; the sample length N represents the length of the time series, and should satisfy N≥1024; the delay time τ has a very small effect on the entropy value, and is generally set to τ=1; the rotation of the embedding dimension m will have a greater impact on the MWPE. If the value of m is too small, the state information contained in the time series after reconstruction is less, and the algorithm will not be able to effectively monitor the dynamic mutation of the time series; on the contrary, if the value of m is too large, the time series will be homogenized, without containing subtle changes in the sequence, and the calculation is cumbersome, and the algorithm running time increases. Usually, the value range of m is 3 to 7. In order to further determine the specific value of the embedding dimension m, the coefficient of variation method (CV for short) is introduced for analysis and determination.

[0069] Therefore, the scale factor s is set to ≥ 10, the sample length N should satisfy N ≥ 1024, and the delay time τ = 1;

[0070] The coefficient of variation method is introduced to analyze and determine the embedding dimension m: a set of data is randomly selected from the gear vibration acceleration data of the healthy gear and the faulty gear. In this embodiment, a set of sample data is randomly selected from the healthy gear and the missing tooth gear data. The embedding dimension m = 3, 4, 5, 6, 7 is taken, and the parameters are respectively brought in and the MWPE value of each set of data is calculated. The two gear state signals are as follows Figure 3 As shown in the figure. When the embedding dimension m is 3, 4, and 5, the MWPE entropy values ​​of the two gear states are generally large, with more overlaps and stronger randomness, which cannot well reflect the state information characteristics of the gears; when the embedding dimension m is 7, the entropy value is generally small, the stability is strong, the time series of the phase space reconstruction vector is homogenized, and the subtle changes of the gear signal are ignored, resulting in the algorithm being unable to effectively distinguish the fault type.

[0071] Then the standard deviation and average value of the MWPE values ​​of the two gear data are calculated, and the coefficient of variation CV is calculated using the standard deviation and average value, where CV = standard deviation / average value; the coefficient of variation can measure the degree of dispersion of the entropy value. The results are shown in Table 2. When the embedding dimension m is 6, the coefficient of variation in the two gear states is the smallest, that is, the degree of dispersion of the entropy value is small, the stability is good, and it is easy to characterize the gear state.

[0072] Therefore, in this embodiment, the embedding dimension m=6 is selected.

[0073]

[0074] Table 2 Entropy analysis of two gear signals

[0075] Step 3: Calculate the MWPE value of the data, classify and label the data with the calculated MWPE value according to the fault type, and divide the data of each fault type into training samples and test samples;

[0076] Exemplarily, the MWPE algorithm is based on weighted permutation entropy (WPE). First, the WPE value of the data is calculated. The specific algorithm is as follows:

[0077] For a given signal sequence Multiple subsequences are reconstructed using phase space reconstruction theory. The formula is as follows:

[0078]

[0079] Where m is the embedding dimension and τ is the delay time;

[0080] In order to introduce the amplification characteristics of the signal, each subsequence is calculated The weight value is as follows:

[0081]

[0082] Among them, ω i is the weight value, is the mean of each subsequence, x i+(t-1)τ is a single data point for each subsequence;

[0083] For each signal subsequence Use the weight value ω i and permutation and combination patterns Represents characteristic information; there are K kinds of signal sequence arrangement modes, and the probability weighted value of each mode is:

[0084]

[0085] in, for The arrangement mode N is the given signal length, is a subsequence of each signal;

[0086] Calculate the WPE value of the signal sequence, the formula is as follows:

[0087]

[0088] in, The probability weighted value for each mode, is a subsequence of each signal, m is the embedding dimension, and τ is the delay time;

[0089] WPE is a single scale to measure the complexity and dynamics of data, while MWPE coarsens the time series and then calculates the WPE value at each scale, which is a quantitative description of the time series at different sizes.

[0090] For a given signal sequence x with a length of N i The coarse-grained sequence is obtained by coarse-grained processing, and the formula is as follows:

[0091]

[0092] Among them, y (s) (j) is the coarse-grained sequence, s is the scale factor, and N is the given signal length;

[0093] Finally, the coarse-grained sequence is used to calculate the WPE value at multiple scales, that is, the MWPE value, and the formula is as follows:

[0094] MWPE(X,τ,s,m)=WPE(y s (j),m,τ)

[0095] In order to facilitate the analysis of the calculated entropy value in the same dimension, it is normalized. The formula is as follows:

[0096]

[0097] The data with calculated MWPE values ​​are classified and labeled according to the fault type, and the data of each fault type are divided into training samples and test samples in a ratio of 4:6.

[0098] Step 4: Use the training samples to train the DBO-SVM classifier model;

[0099] Exemplary, specifically:

[0100] Step S1: input training samples, initialize SVM and DBO parameters, use mapminmax function to normalize sample data to between 0 and 1, use the minimum SVM model recognition error rate as the fitness function, and set the maximum number of DBO iterations to T;

[0101] Step S2: updating the position and fitness function according to the updating rules of different dung beetles. If the new position is better than the original position, the new position replaces the original position, otherwise the original position remains unchanged.

[0102] Step S3: Determine whether the maximum number of iterations is met. If not, return to step S2; if so, output the penalty factor c and kernel function parameter g of the SVM model, which are the optimal parameters.

[0103] Step S4: Use the optimal parameters of the SVM penalty factor c and the kernel function parameter g to train the SVM model.

[0104] Step 5: Use the trained DBO-SVM classifier model to identify the fault type of the test sample;

[0105] Exemplarily, the test sample is input into the DBO-SVM classifier model, the DBO-SVM classifier model processes the data in the test sample and matches it with the fault type, and outputs the corresponding fault type.

[0106] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A gear vibration fault diagnosis method, characterized in that: The method comprises the following steps: Step 1: Use a vibration acceleration sensor to collect gear vibration acceleration data of healthy gears and gears with different faults; Step 2: Analyze the collected data and use the coefficient of variation to determine the parameters in the MWPE algorithm; Step 3: Calculate the MWPE value of the data, classify and label the data with the calculated MWPE value according to the fault type, and divide the data of each fault type into training samples and test samples; Step 4: Use the training samples to train the DBO-SVM classifier model; Step 5: Use the trained DBO-SVM classifier model to identify the fault type of the test sample.

2. A gear vibration fault diagnosis method according to claim 1, characterized in that: The faulty gears described in step 1 include: missing tooth gears, cracked gears, peeled gears, and sharpened gears.

3. A gear vibration fault diagnosis method according to claim 1, characterized in that: The parameters in the MWPE algorithm in step 2 include: sample length N, embedding dimension m, scale factor s, and delay time τ.

4. A gear vibration fault diagnosis method according to claim 3, characterized in that: In step 2, the parameters in the MWPE algorithm are determined, specifically: Set the scale factor s≥10, the sample length N should satisfy N≥1024, and the delay time τ=1; A group of data is randomly selected from the gear vibration acceleration data of healthy gears and faulty gears, and the embedding dimension m is taken as 3, 4, 5, 6, and 7. The parameters are respectively brought in and the MWPE value of each group of data is calculated. Then the standard deviation and average value of the MWPE values ​​of the two gear data are calculated. The coefficient of variation CV is calculated using the standard deviation and average value. The embedding dimension m when the coefficient of variation CV is the minimum value is selected as the final embedding dimension m value; the coefficient of variation CV calculation formula is: Among them, σ is the standard deviation and μ is the mean.

5. A gear vibration fault diagnosis method according to claim 4, characterized in that: The MWPE value of the calculated data in step 3 is: First, calculate the WPE value of the data. The specific algorithm is as follows: For a given signal sequence Multiple subsequences are reconstructed using phase space reconstruction theory. The formula is as follows: Where m is the embedding dimension and τ is the delay time; Calculate each subsequence The weight value is as follows: Among them, ω i is the weight value, is the mean of each subsequence, x i+(t-1)τ is a single data point for each subsequence; For each signal subsequence Use the weight value ω i and permutation and combination patterns Represents characteristic information; there are K kinds of signal sequence arrangement modes, and the probability weighted value of each mode is: in, for The arrangement mode N is the given signal length, is a subsequence of each signal; Calculate the WPE value of the signal sequence, the formula is as follows: in, is the probability weight for each mode, is a subsequence of each signal, m is the embedding dimension, and τ is the delay time; Then, for the signal sequence x i The coarse-grained sequence is obtained by coarse-grained processing, and the formula is as follows: Where s is the scale factor and N is the given signal length; Finally, the coarse-grained sequence is used to calculate the WPE value at multiple scales, that is, the MWPE value, and the formula is as follows: MWPE(X,τ,s,m)=WPE(y s (j),m,t) Normalization is performed, and the formula is as follows:

6. A gear vibration fault diagnosis method according to claim 1, characterized in that: In step 3, the data of each fault type are divided into training samples and test samples, specifically: the data of each fault type are divided into training samples and test samples in a ratio of 4:

6.

7. A gear vibration fault diagnosis method according to claim 1, characterized in that: In step 4, the DBO-SVM classifier model is trained using training samples, specifically: The training samples are input into the SVM model, and the fitness function is set as the minimum SVM recognition error rate. The optimal parameters of the SVM penalty factor c and kernel function parameter g are selected through the dung beetle algorithm, and the SVM model is trained using the selected optimal parameters.

8. A gear vibration fault diagnosis method according to claim 7, characterized in that: The specific steps are: Step S1: input training samples, initialize SVM and DBO parameters, use mapminmax function to normalize sample data to between 0 and 1, use the minimum SVM model recognition error rate as the fitness function, and set the maximum number of DBO iterations to T; Step S2: updating the position and fitness function according to the updating rules of different dung beetles. If the new position is better than the original position, the new position replaces the original position, otherwise the original position remains unchanged. Step S3: Determine whether the maximum number of iterations is met. If not, return to step S2; if so, output the penalty factor c and kernel function parameter g of the SVM model, which are the optimal parameters. Step S4: Use the optimal parameters of the penalty factor c and kernel function parameter g of the SVM model to train the SVM model.

9. A gear vibration fault diagnosis method according to claim 1, characterized in that: Step 5 is specifically as follows: input the test sample into the DBO-SVM classifier model, the DBO-SVM classifier model processes the data in the test sample and matches it with the fault type, and outputs the corresponding fault type.

Images