Engine vibration source identification method based on adaptive feature extraction

Through the adaptive feature extraction method, combined with the correlation between the reconstructed signal and the original signal and the fourth-order cumulative amount optimization, the problem of underdecomposition and insufficient robustness in the engine vibration signal feature extraction is solved, and the accurate identification of the engine vibration source is achieved. and high-precision extraction in strong interference environments.

CN119000099APending Publication Date: 2024-11-22CHINA NORTH ENGINE RES INST
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410907589.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-08
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The prior art has underdecomposition, overdecomposition and endpoint effects in engine vibration signal feature extraction, resulting in signal feature distortion, and insufficient robustness in a strong interference environment, making it difficult to achieve accurate vibration source identification.

Method used

A method for identifying engine vibration source based on adaptive feature extraction is proposed. Through the whole machine vibration testing, detrend processing, adaptive decomposition, denoising processing, time-frequency analysis and order spectrum feature extraction, the reconstruction decomposition signal and original signal are used. The correlation of the signal and fourth-order cumulative amount are the VME optimization judgment indicators, and optimized variational modal extraction (OVME) is realized to improve extraction accuracy and robustness.

Benefits of technology

It effectively avoids distortion of signal characteristics, improves the extraction accuracy in strong interference environments, ensures accurate identification of engine vibration sources, and provides higher applicability and reliability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119000099A_ABST
    Figure CN119000099A_ABST
Patent Text Reader

Abstract

The invention provides an engine vibration source identification method based on adaptive feature extraction, and the method comprises the following steps: S1, carrying out the vibration test of a whole machine, and collecting a vibration signal of an internal combustion engine; s2, performing detrending processing on the vibration signal, wherein the processed vibration signal is X (t); s3, self-adaptive decomposition is carried out on the vibration signal X (t); S4, the vibration decomposition component obtained in the step S3 is reconstructed, and denoising processing is carried out on the vibration signal; and S5, carrying out time-frequency analysis and order spectrum feature extraction on the vibration reconstruction signal obtained in the step S4, and carrying out vibration source identification according to an engine working principle and excitation source features. The method has the beneficial effects that the correlation between the reconstructed decomposition signal and the original signal is taken as a constraint condition of VME optimization, and effective characteristics of the vibration signal are decomposed.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of engine vibration noise, and in particular relates to an engine vibration source identification method based on adaptive feature extraction. Background Art

[0002] Engine vibration signals have nonlinear, non-steady-state, strong coupling, and strong interference characteristics. It is difficult to quickly and accurately identify the vibration source, and effective vibration source identification is of great significance to engine vibration reduction, noise reduction, and reliability development. Signal decomposition can decompose the complex vibration signal characteristics of the engine into a series of characteristic components. Accurate vibration feature extraction is the key to engine vibration source identification.

[0003] In 2018, variational mode extraction (VME) was developed based on VMD, and achieved good application results in the medical field and bearing fault diagnosis field for rapid feature extraction. VME has higher extraction accuracy and lower computational cost, but the VME algorithm parameter setting is somewhat subjective, and the signal processing results still have under-decomposition, over-decomposition and endpoint effects, resulting in distortion of decomposed signal features and inability to accurately decompose composite signals. In addition, the engine vibration signal is highly interfering, resulting in insufficient robustness of the VME algorithm in a strong interference environment.

[0004] The engine vibration signal characteristics are highly interfering, impactful, and complex in spectrum composition. It is difficult to extract the signal characteristics with high accuracy and adaptability. Therefore, an adaptive signal feature extraction method suitable for engine vibration signal characteristics is proposed, which is of great significance for the identification of engine vibration sources. Summary of the invention

[0005] In view of this, the present invention aims to propose an engine vibration source identification method based on adaptive feature extraction to solve at least one technical problem in the background technology.

[0006] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0007] A method for identifying engine vibration sources based on adaptive feature extraction comprises the following steps:

[0008] S1: Perform whole machine vibration test and collect internal combustion engine vibration signals;

[0009] S2: Detrending the vibration signal, the processed vibration signal is X(t);

[0010] S3: Adaptive decomposition of vibration signal X(t):

[0011] S4: reconstructing the vibration decomposition components obtained in step S3 and performing denoising on the vibration signal;

[0012] S5: Perform time-frequency analysis and order spectrum feature extraction on the vibration reconstruction signal obtained in step S4, and identify the vibration source according to the engine working principle and excitation source characteristics.

[0013] Furthermore, the adaptive decomposition of the vibration signal X(t) in step S3 includes the following steps:

[0014] A1: Initialize the decomposition component K, K = 1;

[0015] A2: Perform signal VME decomposition on the vibration signal preprocessed in step S2 to obtain decomposition components IMF(t);

[0016] A3: Calculate the correlation coefficient Corr(K) between the reconstructed signal Sum_IMF(t) of the decomposed component and the original signal X(t);

[0017] A4: If the number of decomposition levels K ≥ 2, execute step A5; otherwise, K = K + 1, and execute steps A2-A4 repeatedly;

[0018] A5: If Corr X,Y >0.95, the vibration signal decomposition result C(t) is output, otherwise K=K+1, and steps A2-A5 are executed cyclically;

[0019] A6: Calculate the fourth-order cumulative quantity P(i) of the vibration decomposition component C(t) output in step A4 according to the high-order cumulative quantity theory;

[0020] A7: Calculate the quotient of adjacent fourth-order cumulants Ratio_P(i);

[0021] A8: If Ratio_P(i)<10, loop through steps A7 and A8; otherwise, output the optimal decomposition level to complete the OVME decomposition of the engine vibration signal.

[0022] Furthermore, in step A3, Sum_IMF(t) represents the sum of the matrix columns of the decomposed component IMF(t), and the Pearson correlation coefficient between the reconstructed signal Sum_IMF(t) and the original signal is calculated by the following calculation formula:

[0023]

[0024] X represents the preprocessed original vibration signal;

[0025] Y represents the sum signal Sum_IMF of the VME decomposition component IMF;

[0026] K represents the number of decomposition levels;

[0027] E represents the expectation of the signal data.

[0028] Furthermore, in step A7, the quotient of adjacent fourth-order cumulants Ratio_P(i), i=1, 2, 3…K-1 integers, is calculated as follows:

[0029] Ratio_P(i)=P(i) / P(i+1); (2)

[0030] Ratio_P(i) is the quotient of adjacent fourth-order cumulants;

[0031] P(i) is the fourth-order cumulant of i;

[0032] P(i+1) is the fourth-order cumulant of i+1.

[0033] Furthermore, the denoising process in step S4 is to reconstruct the vibration decomposition components obtained in step S3 to ensure that the mutual information between the reconstructed signal and the original signal exceeds 0.8. The mutual information can characterize the correlation of information features. The mutual information calculation is shown in formula (12), where: USum_IMF represents the reconstructed signal of the decomposed component, X represents the original signal; I represents the mutual information between the reconstructed signal and the original vibration signal.

[0034]

[0035] U Sum_IMF a reconstructed signal representing the decomposed components;

[0036] X represents the original signal.

[0037] Furthermore, step S5 also includes performing time-frequency analysis and order spectrum feature extraction on the vibration reconstruction signal obtained in step S4, extracting the time-frequency features of the reconstructed signal using wavelet transform, analyzing the spectrum distribution of the vibration signal and the corresponding time domain information, and extracting the order spectrum features of the vibration signal according to the relationship between the frequency domain and the order domain. Based on the working principle of the engine, the time-frequency analysis, and the order spectrum analysis results, the main vibration source of the engine is identified.

[0038] Compared with the prior art, the engine vibration source identification method based on adaptive feature extraction described in the present invention has the following advantages:

[0039] 1. This application uses the correlation between the reconstructed decomposed signal and the original signal as the constraint condition for VME optimization, ensuring that the effective features of the vibration signal are decomposed.

[0040] 2. This application uses the fourth-order cumulant as the judgment indicator for VME optimization and proposes a vibration signal decomposition method based on optimized modal extraction. This method is still highly applicable to strong impact vibration characteristics and provides guidance for the identification of engine vibration sources under strong impact environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0042] Figure 1 A schematic diagram of an engine vibration source identification method based on adaptive feature extraction according to an embodiment of the present invention;

[0043] Figure 2 A schematic diagram of adaptive decomposition of vibration signals based on OVME according to an embodiment of the present invention;

[0044] Figure 3 Schematic diagram of time-frequency analysis and order spectrum analysis of vibration signals (a is time-frequency analysis, b is time-frequency analysis). DETAILED DESCRIPTION

[0045] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0046] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0047] Aiming at the problem that the variational mode extraction (VME) method has insufficient accuracy and robustness in processing engine vibration signals, this patent takes the correlation between the reconstructed decomposition signal and the original signal as the constraint condition and the fourth-order cumulant as the judgment index for VME optimization, and proposes an optimized variational mode extraction method (OptimizedVariationalModeExtraction, OVME) to provide guidance for engine vibration feature extraction and vibration source identification.

[0048] (1) Engine bench test

[0049] The vibration test was carried out on the typical test points of the engine. The test system consists of LMS acquisition front end, three-axis acceleration sensor (PCB, USA), LMS Test.Lab test and post-processing software and laptop computer. The acquisition frequency of the vibration signal is 21500Hz, the frequency resolution is 1Hz, and the test time is 8s.

[0050] (2) Vibration signal preprocessing

[0051] The vibration signal collected in step (1) contains instrument collection errors, and the vibration signal is detrended. The processed vibration signal is X(t), which provides input for subsequent engine vibration signal characteristics.

[0052] (3) Adaptive decomposition of vibration signals

[0053] Taking the correlation between the reconstructed decomposition signal and the original signal as the constraint condition and the fourth-order cumulant as the judgment index of VME optimization, a vibration signal decomposition method of optimized variational mode extraction (OVME) is proposed. The specific steps are as follows:

[0054] 1) Initialize the decomposition component K, K = 1;

[0055] 2) performing signal VME decomposition on the vibration signal preprocessed in step (2) to obtain decomposition component IMF(t);

[0056] 3) Calculate the correlation coefficient Corr(K) between the reconstructed signal Sum_IMF(t) of the decomposition component and the original signal X(t); Sum_IMF(t) represents the sum of the matrix columns of the decomposition component IMF(t), and the Pearson correlation coefficient between the reconstructed signal Sum_IMF(t) and the original signal is calculated as follows:

[0057]

[0058]

[0059] Where X represents the preprocessed original vibration signal, Y represents the sum signal Sum_IMF of the VME decomposition component IMF, K represents the number of decomposition layers, and E represents the expectation of the signal data;

[0060] 4) If the number of decomposition levels K ≥ 2, proceed to the next step, otherwise K = K + 1 and loop steps 2) - 4);

[0061] 5) If Corr X,Y >0.95, exit the loop and output the vibration signal decomposition result C(t). Otherwise, K=K+1 and repeat steps 2)-5);

[0062] 6) According to the high-order cumulative quantity theory, the fourth-order cumulative quantity P(i) of the vibration decomposition component C(t) output in step (4) is calculated, where i represents the number of decomposition levels. The calculation principle of the fourth-order cumulative quantity P(i) is as follows:

[0063] For a continuous random variable x, its probability density function is f(x). In this patent, x is the vibration signal decomposition component C(t). Its first characteristic function can be obtained by Laplace transform, as shown in formula (2). Its k-order derivative can be expressed as formula (3):

[0064]

[0065] Let ω = 0, then the k-th moment m of the variable x isk It is expressed as formula (4):

[0066] m k =Φ k (0) = E{x k} (4)

[0067] The first characteristic function Φ(ω) is also called the moment generating function. Based on this, the second characteristic function, namely the cumulant generating function, can be obtained, as shown in the following formula (5):

[0068] Ψ(ω)=1nΦ(ω) (5)

[0069] The value of the k-order derivative of the function Ψ(ω) at ω = 0 is the k-order cumulant c of the random variable x. k :

[0070]

[0071] From formula (2), we know that Φ(0) = 1. Substituting this result into formula (4), we get ψ(0) = 0. Ψ(ω) can be expanded using Taylor series as shown below:

[0072]

[0073]

[0074] If ω=0, we can get:

[0075]

[0076] Formula (6) and formula (9) can be combined to obtain: c 1 =m 1 ;

[0077] c1 and c2 are the first-order and second-order cumulants, respectively. m1 and m2 are the first-order and second-order moments corresponding to equation (4). Similarly, higher-order cumulants can be calculated. The r = 4th-order cumulant P(i) of the random variable C(t) can be defined as:

[0078]

[0079] 7) Calculate the quotient of adjacent fourth-order cumulants Ratio_P(i), where i=1, 2, 3…K-1 integers, and the specific calculation is as follows:

[0080] Ratio_P(i)=P(i) / P(i+1) (11)

[0081] 8) If Ratio_P(i) < 10, loop step 7) and step 8), otherwise exit the loop, output the optimal decomposition layer number, and complete the OVME decomposition of the engine vibration signal.

[0082] (4) Engine vibration signal denoising

[0083] In order to enhance the characteristics of the vibration signal, it is necessary to reconstruct the decomposed components. The vibration decomposition components obtained in step (3) are reconstructed to ensure that the mutual information between the reconstructed signal and the original signal exceeds 0.8 to achieve the denoising of the vibration signal. Mutual information can characterize the correlation of information features. The calculation of mutual information is shown in formula (12), where U Sum_IMF represents the reconstructed signal of the decomposed components, X represents the original signal, that is, the vibration signal X(t) preprocessed in step (2), and I represents the mutual information between the reconstructed signal and the original vibration signal.

[0084]

[0085] (5) Engine vibration source identification

[0086] Perform time-frequency analysis and order spectrum feature extraction on the vibration reconstruction signal obtained in step (4). Use wavelet transform to extract the time-frequency features of the reconstructed signal, and analyze the spectrum distribution of the vibration signal and the corresponding time domain information. According to the relationship between the frequency domain and the order domain, extract the order spectrum features of the vibration signal. Based on the working principle of the engine, time-frequency analysis, and order spectrum analysis results, identify the main vibration source of the engine.

[0087] (6) Application of engine vibration source identification method based on adaptive feature extraction

[0088] Based on the engine test bench, the vibration test of the engine body and the measuring points near the transmission box in step (1) is completed, and the vibration signal is preprocessed using the method in step (2). The vibration signal is preprocessed based on the OVME method proposed in step (3) for decomposition. The decomposition result is as follows: Figure 2 As shown, Figure 2 It can be seen that the vibration signal is decomposed into 6 decomposition components. The mutual information between the reconstructed vibration component and the original signal is calculated using the method in step (4) to complete the vibration signal denoising. Step (5) is used to perform time-frequency analysis of the reconstructed vibration signal ( Figure 3 a) and order spectrum feature extraction ( Figure 3 b), Figure 3 a It can be seen that the vibration signal is mainly distributed in 3000-3800Hz. Figure 3 b It can be seen that the main energy of the vibration signal is distributed in the 54th order. According to the structural characteristics of the engine and the order characteristics of the transmission load, it is found that the vibration mainly comes from the second-order meshing load of the gear, and the vibration source is the meshing excitation of the gear.

[0089] This patent proposes an adaptive vibration signal decomposition method suitable for engine characteristics, which combines vibration signal time-frequency analysis and order spectrum analysis to achieve effective identification of engine vibration sources.

[0090] 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. An engine vibration source identification method based on adaptive feature extraction, Features: The steps include: S1: Perform whole machine vibration test and collect internal combustion engine vibration signals; S2: Detrending the vibration signal, the processed vibration signal is X(t); S3: Adaptive decomposition of vibration signal X(t): S4: reconstructing the vibration decomposition components obtained in step S3 and performing denoising on the vibration signal; S5: Perform time-frequency analysis and order spectrum feature extraction on the vibration reconstruction signal obtained in step S4, and identify the vibration source according to the engine working principle and excitation source characteristics.

2. The engine vibration source identification method based on adaptive feature extraction according to claim 1, Features: The adaptive decomposition of the vibration signal X(t) in step S3 includes the following steps: A1: Initialize the decomposition component K, K = 1; A2: Perform signal VME decomposition on the vibration signal preprocessed in step S2 to obtain decomposition components IMF(t); A3: Calculate the correlation coefficient Corr(K) between the reconstructed signal Sum_IMF(t) of the decomposed component and the original signal X(t); A4: If the number of decomposition levels K ≥ 2, execute step A5; otherwise, K = K + 1, and execute steps A2-A4 repeatedly; A5: If Corr X,Y >0.95, the vibration signal decomposition result C(t) is output, otherwise K=K+1, and steps A2-A5 are executed cyclically; A6: Calculate the fourth-order cumulative quantity P(i) of the vibration decomposition component C(t) output in step A4 according to the high-order cumulative quantity theory; A7: Calculate the quotient of adjacent fourth-order cumulants Ratio_P(i); A8: If Ratio_P(i)<10, loop through steps A7 and A8; otherwise, output the optimal decomposition level to complete the OVME decomposition of the engine vibration signal.

3. The engine vibration source identification method based on adaptive feature extraction according to claim 2, Features: In step A3, Sum_IMF(t) represents the sum of the matrix columns of the decomposed component IMF(t). The Pearson correlation coefficient between the reconstructed signal Sum_IMF(t) and the original signal is calculated using the following formula: X represents the preprocessed original vibration signal; Y represents the sum signal Sum_IMF of the VME decomposition component IMF; K represents the number of decomposition levels; E represents the expectation of the signal data.

4. The engine vibration source identification method based on adaptive feature extraction according to claim 2, Features: In step A7, the quotient of adjacent fourth-order cumulants Ratio_P(i), i=1, 2, 3…K-1 integer, is calculated as follows: Ratio_P(i)=P(i) / P(i+1); (2) Ratio_P(i) is the quotient of adjacent fourth-order cumulants; P(i) is the fourth-order cumulant of i; P(i+1) is the fourth-order cumulant of i+1.

5. The engine vibration source identification method based on adaptive feature extraction according to claim 1, Features: The denoising process in step S4 is to reconstruct the vibration decomposition components obtained in step S3 to ensure that the mutual information between the reconstructed signal and the original signal exceeds 0.

8. The mutual information can characterize the correlation of information features. The mutual information calculation is shown in formula (12), where: USum_IMF represents the reconstructed signal of the decomposed component, X represents the original signal; I represents the mutual information between the reconstructed signal and the original vibration signal; USum_IMF a reconstructed signal representing the decomposed components; X represents the original signal.

6. The engine vibration source identification method based on adaptive feature extraction according to claim 1, Features: Step S5 also includes performing time-frequency analysis and order spectrum feature extraction on the vibration reconstruction signal obtained in step S4, extracting the time-frequency features of the reconstructed signal using wavelet transform, analyzing the spectrum distribution of the vibration signal and the corresponding time domain information, and extracting the order spectrum features of the vibration signal according to the relationship between the frequency domain and the order domain; Based on the engine working principle, time-frequency analysis and order spectrum analysis results, the main vibration sources of the engine are identified.

Citation Information

Cited By

  • Method and system for reconstructing timing undersampling vibration signal of rotating blade of aero-engine

    CN120145091A