Method based on vibration signal frequency extraction algorithm

Through signal acquisition, preprocessing, time domain analysis and frequency domain conversion, combined with deep learning and machine learning technology, the vibration signal frequency extraction algorithm is optimized, which solves the problem of poor frequency extraction accuracy in the existing technology, and improves the diagnosis of vibration signals and Monitor accuracy to ensure system stability.

CN118964967APending Publication Date: 2024-11-15ZHONGXUN DIGITAL TECHNOLOGY (SUZHOU) CO LTD

Patent Information

Application Number
CN202411054490.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-02
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

Existing vibration signal frequency extraction algorithms are difficult to accurately capture nonlinear and time-varying vibration signal characteristics, resulting in poor frequency extraction accuracy, affecting the accuracy of fault diagnosis and mechanical state monitoring. Especially in systems that require precise control, the frequency extraction accuracy is low. Will reduce system performance.

Method used

Using methods such as signal acquisition and preprocessing, time domain analysis, frequency domain conversion and frequency extraction, including filtering, denoising, smoothing, Fourier transform, feature parameter extraction and model verification, combined with deep learning and machine learning technology, optimized Frequency extraction algorithms to adapt to different types of vibration signals.

Benefits of technology

It improves the accuracy and robustness of vibration signal frequency extraction, reduces the impact of noise, enhances the accuracy of diagnosis and monitoring, and ensures the stability and control effect of the system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118964967A_ABST
    Figure CN118964967A_ABST
Patent Text Reader

Abstract

The invention discloses a method based on a vibration signal frequency extraction algorithm, and relates to the technical field of signal processing, and the method comprises the following steps: S1, signal collection and preprocessing; s2, carrying out time domain analysis; s3, performing frequency domain conversion; s4, frequency extraction; and S5, verifying and optimizing a result. According to the method based on the vibration signal frequency extraction algorithm, signal preprocessing is optimized, the signal-to-noise ratio of signals is increased and the influence of noise on frequency extraction is reduced by improving preprocessing steps of denoising, filtering and the like, a proper frequency extraction algorithm is selected for different types of vibration signals, and parameter optimization and performance adjustment are carried out. The advanced signal processing technology is introduced, the modern signal processing technology such as deep learning and machine learning is combined, the accuracy and robustness of frequency extraction are improved, the improved method comprehensively considers the aspects of signal characteristics, algorithm selection and the signal processing technology, and the precision of vibration signal frequency extraction is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of signal processing, and in particular to a method based on a vibration signal frequency extraction algorithm. Background Art

[0002] With the continuous improvement of sensor technology and the substantial improvement of computer performance, it has become possible to accurately and efficiently collect, process and analyze vibration signals. The frequency extraction algorithm based on vibration signals refers to extracting key frequency components from vibration signals through a series of mathematical and signal processing techniques. The purpose of this algorithm is to conduct in-depth analysis of the vibration characteristics of mechanical equipment or structures, so as to identify the status of the equipment, failure mode or other relevant information.

[0003] For example, the patent document with application number 202310247110.7 discloses a method based on a vibration signal frequency interval extraction algorithm. Compared with the prior art, this method improves the dependence on professional technicians and professional knowledge in the traditional diagnosis process and makes the diagnosis technology simpler. However, the existing method based on the vibration signal frequency extraction algorithm similar to the above application still has the following shortcomings:

[0004] In the prior art, since vibration signals usually have nonlinear and time-varying characteristics, which means that the characteristics of the signal will change over time, and traditional frequency extraction algorithms are difficult to accurately capture these changes, and different frequency extraction algorithms have different applicable scopes and performance characteristics. They are only effective for specific types of vibration signals, but may perform poorly for other types of signals, thereby affecting the frequency extraction effect and resulting in poor accuracy in vibration signal frequency extraction. In fields such as fault diagnosis or mechanical condition monitoring, the accuracy of vibration signal frequency extraction directly affects the accuracy of diagnosis or monitoring, affecting maintenance decisions or safety assessments. In addition, the frequency extraction accuracy also directly affects the subsequent processing of vibration signals. If the frequency extraction is inaccurate, the results of subsequent processing will also be biased. Especially in some systems that require precise control or adjustment, the low accuracy of frequency extraction will reduce system performance, causing the system to fail to achieve the expected control effect or operational stability.

[0005] Therefore, it is urgent to improve this shortcoming. The present invention studies and improves the existing structure and shortcomings, and provides a method based on a vibration signal frequency extraction algorithm. Summary of the invention

[0006] The purpose of the present invention is to provide a method based on a vibration signal frequency extraction algorithm to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solution: a method based on a vibration signal frequency extraction algorithm, comprising the following steps:

[0008] S1. Signal acquisition and preprocessing: Vibration signals are collected through sensors and data collectors. The vibration signals come from various mechanical equipment or systems, such as engines, bearings, gears, etc., and the collected original signals are preprocessed to remove noise interference and unnecessary components;

[0009] S2. Time domain analysis: By statistically analyzing the vibration signal, key information such as the amplitude and period of the signal is extracted to preliminarily judge the working condition of the equipment, identify the periodicity of the signal, and whether there is abnormal vibration such as resonance and beat frequency;

[0010] S3, frequency domain conversion: convert the time domain signal into the frequency domain signal through Fourier transform, obtain the spectrum distribution of the signal, including the amplitude and phase information of each frequency component, analyze the frequency component of the signal, and further understand the frequency characteristics of the signal;

[0011] S4. Frequency extraction: extracting characteristic parameters related to the frequency of the vibration signal in the frequency domain, wherein the characteristic parameters include but are not limited to the main frequency, the secondary frequency, the frequency distribution range, and the amplitude, and applying a specific algorithm or model based on the extracted characteristic parameters to extract the target frequency in the vibration signal;

[0012] S5. Result verification and optimization: Verify and optimize the extracted frequency results by comparing with actual measurement data or other known data.

[0013] Furthermore, in step S1, the preprocessing includes filtering, denoising, and smoothing operations, which are specifically as follows:

[0014] Filtering: Remove noise and interference from vibration signals through hardware filtering or software filtering. Specifically, in software filtering, fast Fourier transform filtering and wavelet analysis filtering are used. The fast Fourier transform filtering is based on the FFT algorithm and can effectively analyze and process the frequency components of the signal. The wavelet analysis filtering uses the characteristics of wavelet transform to perform multi-scale and multi-resolution analysis on the signal to extract useful information hidden in the noise.

[0015] Denoising: Determine the type and distribution characteristics of noise according to the characteristics and requirements of the signal, and use a suitable denoising algorithm to remove the noise components in the signal. The denoising algorithm includes but is not limited to denoising based on statistical models and denoising based on wavelet transform;

[0016] Smoothing: The five-point cubic smoothing method is used to eliminate random fluctuations and local outliers in the signal, making the signal smoother and more stable. The five-point cubic smoothing method takes two adjacent points in front of each data point, approximates it with a cubic polynomial, and determines the coefficients of the polynomial according to the least squares principle to achieve signal smoothing.

[0017] Furthermore, in step S2, the time domain analysis analyzes the characteristics of the signal changing over time by observing the changes of the signal in the time domain, thereby providing basic data for subsequent frequency domain conversion and frequency extraction.

[0018] Furthermore, in step S2, the time domain analysis operation specifically includes the following steps:

[0019] S21. Examine the characteristics of the value sequence, such as periodicity, trend, randomness, etc.;

[0020] S22. Select a suitable fitting model according to the characteristics of the sequence, such as a linear model, an autoregressive model, etc.;

[0021] S23. Determine the parameters of the model based on the observed data, usually involving methods such as least squares and maximum likelihood estimation;

[0022] S24. Test the model and optimize it to ensure that the model can accurately describe the characteristics of the sequence;

[0023] S25. Use the fitted model to infer other statistical properties of the sequence or predict the future development of the sequence.

[0024] Furthermore, in step S3, the Fourier transform analyzes the frequency characteristics of the signal by decomposing the complex time domain signal into a simple sine wave signal, and the formulas involved in the Fourier transform include Fourier series and Fourier integral. For continuous-time signals, the Fourier integral is used for transformation, and for discrete-time signals, the Fourier series expansion is used.

[0025] Furthermore, in step S3, the specific operation of the Fourier transform includes: determining the function to be transformed, and performing periodic extension on the signal, and then selecting an appropriate transformation method according to whether the signal is a continuous-time signal or a discrete-time signal, performing Fourier integration or Fourier series expansion, and after obtaining the frequency domain representation, understanding the frequency domain characteristics of the signal, such as frequency components, energy distribution, etc., by analyzing the signal's spectrum.

[0026] Furthermore, in step S3, the Fourier transform may select whether to perform an inverse transform on the frequency domain representation as required, so as to reconvert the frequency domain signal into a time domain signal for further processing or to restore the original signal.

[0027] Furthermore, in step S4, the specific operation of feature parameter extraction includes: analyzing the correlation between the extracted features and the target frequency, selecting features with strong correlation, and reducing feature dimensions and removing redundant information through principal component analysis and independent component analysis techniques.

[0028] Furthermore, in step S4, the specific operation of the algorithm or model application is as follows:

[0029] Step 1: Target frequency extraction algorithm or model selection:

[0030] Spectrum peak search: Search for peaks on the spectrum graph to determine the location of the target frequency;

[0031] Machine learning model: Use supervised learning algorithms such as regression and classification to train the model so that it can identify and extract target frequencies;

[0032] Deep learning model: Applying neural networks to perform end-to-end learning on signals to identify and extract target frequencies. The neural networks include but are not limited to convolutional neural networks and recurrent neural networks.

[0033] Step 2: Target frequency extraction:

[0034] Threshold-based: Set a threshold, search for frequency components exceeding the threshold on the spectrum graph, and determine them as target frequencies;

[0035] Use model prediction: For machine learning or deep learning models, the preprocessed vibration signal is input and the model outputs the predicted target frequency.

[0036] Furthermore, in step S5, if the result is inaccurate or has errors, the preprocessing step, the frequency domain conversion method or the feature extraction algorithm is adjusted to improve the accuracy and reliability of the frequency extraction.

[0037] The present invention provides a method based on a vibration signal frequency extraction algorithm, which has the following beneficial effects:

[0038] The present invention optimizes signal preprocessing, improves the signal-to-noise ratio of the signal by improving preprocessing steps such as denoising and filtering, reduces the impact of noise on frequency extraction, and selects suitable frequency extraction algorithms for different types of vibration signals, performs parameter optimization and performance adjustment. In addition, advanced signal processing technology is introduced, combined with modern signal processing technology, such as deep learning and machine learning, to improve the accuracy and robustness of frequency extraction. The improved method comprehensively considers aspects such as signal characteristics, algorithm selection and signal processing technology, and improves the accuracy of vibration signal frequency extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic diagram of the process steps of a method based on a vibration signal frequency extraction algorithm of the present invention;

[0040] Figure 2 The present invention is a schematic diagram of the time domain analysis operation of a method based on a vibration signal frequency extraction algorithm. DETAILED DESCRIPTION

[0041] The following embodiments of the present invention are described in further detail in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0042] like Figure 1-Figure 2 As shown, a method based on a vibration signal frequency extraction algorithm comprises the following steps:

[0043] S1. Signal acquisition and preprocessing: Vibration signals are collected through sensors and data collectors. Vibration signals come from various mechanical equipment or systems, such as engines, bearings, gears, etc., and the collected raw signals are preprocessed to remove noise interference and unnecessary components;

[0044] In step S1, preprocessing includes filtering, denoising, and smoothing operations, which are as follows:

[0045] Filtering: Remove noise and interference from vibration signals through hardware filtering or software filtering. Specifically, in software filtering, fast Fourier transform filtering and wavelet analysis filtering are used. Fast Fourier transform filtering is based on the FFT algorithm and can effectively analyze and process the frequency components of the signal. Wavelet analysis filtering uses the characteristics of wavelet transform to perform multi-scale and multi-resolution analysis on the signal to extract useful information hidden in the noise.

[0046] Denoising: Determine the type and distribution characteristics of noise according to the characteristics and requirements of the signal, and use appropriate denoising algorithms to remove the noise components in the signal. Denoising algorithms include but are not limited to denoising based on statistical models and denoising based on wavelet transform;

[0047] Smoothing: The five-point cubic smoothing method is used to eliminate random fluctuations and local outliers in the signal, making the signal smoother and more stable. The five-point cubic smoothing method takes two adjacent points in front of each data point, approximates it with a cubic polynomial, and determines the coefficients of the polynomial according to the least squares principle to achieve signal smoothing.

[0048] S2. Time domain analysis: By statistically analyzing the vibration signal, key information such as the amplitude and period of the signal is extracted to preliminarily judge the working condition of the equipment, identify the periodicity of the signal, and whether there is abnormal vibration such as resonance and beat frequency;

[0049] In step S2, the time domain analysis analyzes the characteristics of the signal changing over time by observing the changes of the signal in the time domain, and provides basic data for subsequent frequency domain conversion and frequency extraction;

[0050] In step S2, the time domain analysis operation specifically includes the following steps:

[0051] S21. Examine the characteristics of the value sequence, such as periodicity, trend, randomness, etc.;

[0052] S22. Select a suitable fitting model according to the characteristics of the sequence, such as a linear model, an autoregressive model, etc.;

[0053] S23. Determine the parameters of the model based on the observed data, usually involving methods such as least squares and maximum likelihood estimation;

[0054] S24. Test the model and optimize it to ensure that the model can accurately describe the characteristics of the sequence;

[0055] S25. Use the fitted model to infer other statistical properties of the sequence or predict the future development of the sequence;

[0056] S3, frequency domain conversion: convert the time domain signal into the frequency domain signal through Fourier transform, obtain the spectrum distribution of the signal, including the amplitude and phase information of each frequency component, analyze the frequency component of the signal, and further understand the frequency characteristics of the signal;

[0057] In step S3, Fourier transform analyzes the frequency characteristics of the signal by decomposing the complex time domain signal into a simple sine wave signal, and the formula involved in Fourier transform includes Fourier series and Fourier integral. For continuous time signals, Fourier integral is used for transformation, and for discrete time signals, Fourier series expansion is used;

[0058] In step S3, the specific operation of Fourier transform includes: determining the function to be transformed, and performing periodic extension on the signal, and then selecting an appropriate transformation method according to whether the signal is a continuous time signal or a discrete time signal, performing Fourier integration or Fourier series expansion, and after obtaining the frequency domain representation, analyzing the signal spectrum to understand the frequency domain characteristics of the signal, such as frequency components, energy distribution, etc.;

[0059] In step S3, the Fourier transform may select whether to perform an inverse transform on the frequency domain representation as required, and reconvert the frequency domain signal into a time domain signal for further processing or to restore the original signal;

[0060] S4. Frequency extraction: In the frequency domain, characteristic parameters related to the frequency of the vibration signal are extracted, and the characteristic parameters include but are not limited to the main frequency, secondary frequency, frequency distribution range, and amplitude. Based on the extracted characteristic parameters, a specific algorithm or model is applied to extract the target frequency in the vibration signal;

[0061] In step S4, the specific operation of feature parameter extraction includes: analyzing the correlation between the extracted features and the target frequency, selecting features with strong correlation, and reducing the feature dimension through principal component analysis and independent component analysis techniques to remove redundant information;

[0062] In step S4, the specific operations of the algorithm or model application are as follows:

[0063] Step 1: Target frequency extraction algorithm or model selection:

[0064] Spectrum peak search: Search for peaks on the spectrum graph to determine the location of the target frequency;

[0065] Machine learning model: Use supervised learning algorithms such as regression and classification to train the model so that it can identify and extract target frequencies;

[0066] Deep learning model: Apply neural networks to perform end-to-end learning on signals to identify and extract target frequencies. Neural networks include but are not limited to convolutional neural networks and recurrent neural networks.

[0067] Step 2: Target frequency extraction:

[0068] Threshold-based: Set a threshold, search for frequency components exceeding the threshold on the spectrum graph, and determine them as target frequencies;

[0069] Use model prediction: For machine learning or deep learning models, the preprocessed vibration signal is input and the model outputs the predicted target frequency;

[0070] S5. Result verification and optimization: Verify and optimize the extracted frequency results by comparing with actual measurement data or other known data;

[0071] In step S5, if the result is inaccurate or has errors, the preprocessing steps, the frequency domain conversion method or the feature extraction algorithm are adjusted to improve the accuracy and reliability of the frequency extraction.

[0072] In summary, combined Figure 1-Figure 2 As shown, the method based on the vibration signal frequency extraction algorithm includes the following steps when used:

[0073] S1. Signal acquisition and preprocessing: Vibration signals are collected through sensors and data collectors. Vibration signals come from various mechanical equipment or systems, such as engines, bearings, gears, etc., and the collected original signals are preprocessed to remove noise interference and unnecessary components. Preprocessing includes filtering, denoising, and smoothing operations. Filtering: Remove noise and interference in vibration signals through hardware filtering or software filtering. Denoising: Determine the type and distribution characteristics of noise according to the characteristics and requirements of the signal, and use a suitable denoising algorithm to remove the noise components in the signal. Smoothing: Use the five-point cubic smoothing method to eliminate random fluctuations and local outliers in the signal.

[0074] S2. Time domain analysis: By statistically analyzing the vibration signal, extracting key information such as the amplitude and period of the signal, preliminarily judging the working condition of the equipment, and identifying the periodicity of the signal and whether there are abnormal vibrations such as resonance and beat frequency; time domain analysis observes the change of the signal in the time domain, analyzes the characteristics of the signal changing over time, and provides basic data for subsequent frequency domain conversion and frequency extraction. The specific operations of time domain analysis include: examining the characteristics of the value sequence, selecting a suitable fitting model based on the characteristics of the sequence, determining the parameters of the model based on the observed data, testing the model and optimizing it, and using the fitted model to infer other statistical properties of the sequence or predict the future development of the sequence;

[0075] S3. Frequency domain conversion: The time domain signal is converted into the frequency domain signal through Fourier transform to obtain the spectrum distribution of the signal, including the amplitude and phase information of each frequency component, analyze the frequency component of the signal, and further understand the frequency characteristics of the signal; wherein, the Fourier transform analyzes the frequency characteristics of the signal by decomposing the complex time domain signal into a simple sine wave signal, and the formulas involved in the Fourier transform include Fourier series and Fourier integral. For continuous time signals, the Fourier integral is used for transformation, and for discrete time signals, the Fourier series expansion is used; the specific operations of the Fourier transform include: determining the function to be transformed, and performing periodic extension on the signal, and then selecting an appropriate transformation method according to whether the signal is a continuous time signal or a discrete time signal, performing Fourier integral or Fourier series expansion, and obtaining the frequency domain representation, and then analyzing the signal spectrum to understand the frequency domain characteristics of the signal; in addition, the Fourier transform can choose whether to perform an inverse transform on the frequency domain representation as needed, and reconvert the frequency domain signal into a time domain signal for further processing or restoration of the original signal;

[0076] S4. Frequency extraction: In the frequency domain, characteristic parameters related to the frequency of the vibration signal are extracted. The characteristic parameters include but are not limited to the main frequency, secondary frequency, frequency distribution range, and amplitude. The specific operations of characteristic parameter extraction include: analyzing the correlation between the extracted features and the target frequency, selecting features with strong correlation, and reducing the feature dimension and removing redundant information through principal component analysis and independent component analysis techniques; and based on the extracted characteristic parameters, applying a specific algorithm or model to extract the target frequency in the vibration signal, and the specific operations of the algorithm or model application are as follows:

[0077] Step 1: Target frequency extraction algorithm or model selection:

[0078] Spectrum peak search: Search for peaks on the spectrum graph to determine the location of the target frequency;

[0079] Machine learning model: Use supervised learning algorithms such as regression and classification to train the model so that it can identify and extract target frequencies;

[0080] Deep learning model: Apply neural networks to perform end-to-end learning on signals to identify and extract target frequencies. Neural networks include but are not limited to convolutional neural networks and recurrent neural networks.

[0081] Step 2: Target frequency extraction:

[0082] Threshold-based: Set a threshold, search for frequency components exceeding the threshold on the spectrum graph, and determine them as target frequencies;

[0083] Use model prediction: For machine learning or deep learning models, the preprocessed vibration signal is input and the model outputs the predicted target frequency;

[0084] S5. Result verification and optimization: Verify and optimize the extracted frequency results by comparing with actual measurement data or other known data. If the results are inaccurate or there are errors, adjust the preprocessing steps, frequency domain conversion method or feature extraction algorithm to improve the accuracy and reliability of frequency extraction.

[0085] The embodiments of the present invention are given for the purpose of illustration and description, and are not intended to be exhaustive or to limit the invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

Claims

1. A method based on vibration signal frequency extraction algorithm, It is characterized in that The following steps are involved: S1. Signal acquisition and preprocessing: Vibration signals are collected through sensors and data collectors, and the collected raw signals are preprocessed to remove noise interference and unnecessary components; S2. Time domain analysis: By statistically analyzing the vibration signal, extract the key information of the signal, preliminarily judge the working condition of the equipment, and identify the periodicity of the signal and whether there is abnormal vibration; S3, frequency domain conversion: convert the time domain signal into the frequency domain signal through Fourier transform, obtain the spectrum distribution of the signal, analyze the frequency components of the signal, and further understand the frequency characteristics of the signal; S4. Frequency extraction: In the frequency domain, characteristic parameters related to the frequency of the vibration signal are extracted, and based on the extracted characteristic parameters, a specific algorithm or model is applied to extract the target frequency in the vibration signal; S5. Result verification and optimization: Verify and optimize the extracted frequency results by comparing with actual measurement data or other known data.

2. A method based on a vibration signal frequency extraction algorithm according to claim 1, It is characterized in that In step S1, the preprocessing includes filtering, denoising, and smoothing operations, which are specifically as follows: Filtering: Remove noise and interference from vibration signals through hardware filtering or software filtering. Specifically, in software filtering, fast Fourier transform filtering and wavelet analysis filtering are used; Denoising: Determine the type and distribution of noise based on the characteristics and requirements of the signal, and use a suitable denoising algorithm to remove the noise components in the signal; Smoothing: The five-point cubic smoothing method is used to eliminate random fluctuations and local outliers in the signal, making the signal smoother and more stable. The five-point cubic smoothing method takes two adjacent points in front of each data point, approximates it with a cubic polynomial, and determines the coefficients of the polynomial according to the least squares principle to achieve signal smoothing.

3. A method based on a vibration signal frequency extraction algorithm according to claim 1, It is characterized in that In step S2, the time domain analysis analyzes the characteristics of the signal changing over time by observing the changes of the signal in the time domain, so as to provide basic data for the subsequent frequency domain conversion and frequency extraction.

4. A method based on a vibration signal frequency extraction algorithm according to claim 1, It is characterized in that In step S2, the time domain analysis operation specifically includes the following steps: S21, examine the characteristics of the value sequence; S22, selecting a suitable fitting model according to the characteristics of the sequence; S23, determining parameters of the model based on the observed data; S24. Test the model and optimize it to ensure that the model can accurately describe the characteristics of the sequence; S25. Use the fitted model to infer other statistical properties of the sequence or predict the future development of the sequence.

5. The method according to claim 1, It is characterized in that In step S3, Fourier transform analyzes the frequency characteristics of the signal by decomposing the complex time domain signal into a simple sine wave signal, and the formulas involved in Fourier transform include Fourier series and Fourier integral. For continuous time signals, Fourier integral is used for transformation, and for discrete time signals, Fourier series expansion is used.

6. A method based on a vibration signal frequency extraction algorithm according to claim 5, It is characterized in that In step S3, the specific operation of Fourier transform includes: determining the function to be transformed, and periodically extending the signal, and then selecting an appropriate transformation method according to whether the signal is a continuous-time signal or a discrete-time signal, performing Fourier integration or Fourier series expansion, and after obtaining the frequency domain representation, understanding the frequency domain characteristics of the signal by analyzing the signal's spectrum.

7. A method based on a vibration signal frequency extraction algorithm according to claim 6, It is characterized in that In step S3, the Fourier transform may select whether to perform an inverse transform on the frequency domain representation as required, so as to reconvert the frequency domain signal into a time domain signal for further processing or to restore the original signal.

8. The method according to claim 1, It is characterized in that In step S4, the specific operation of feature parameter extraction includes: analyzing the correlation between the extracted features and the target frequency, selecting features with strong correlation, and reducing feature dimensions and removing redundant information through principal component analysis and independent component analysis techniques.

9. The method according to claim 1, It is characterized in that In step S4, the specific operation of the algorithm or model application is as follows: Step 1: Target frequency extraction algorithm or model selection: Spectrum peak search: Search for peaks on the spectrum graph to determine the location of the target frequency; Machine learning model: Use supervised learning algorithms to train the model so that it can recognize and extract target frequencies; Deep learning model: Apply neural networks to learn signals end-to-end to identify and extract target frequencies; Step 2: Target frequency extraction: Threshold-based: Set a threshold, search for frequency components exceeding the threshold on the spectrum graph, and determine them as target frequencies; Use model prediction: For machine learning or deep learning models, the preprocessed vibration signal is input and the model outputs the predicted target frequency.

10. The method according to claim 1, It is characterized in that In step S5, if the result is inaccurate or has errors, the preprocessing step, the frequency domain conversion method or the feature extraction algorithm is adjusted to improve the accuracy and reliability of frequency extraction.

Citation Information

Patent Citations

  • Method for extracting fault characteristic frequencies of train rolling bearings with variable rotational speeds

    CN104568444A

  • Bearing fault diagnosis method based on parameter adaptive VMD and optimized SVM

    CN113642508A

  • Method for diagnosing early weak fault signal features of marine machinery

    WO2022165737A1

Cited By

  • Magnetic particle rotation frequency measuring method based on wavelet transformation

    CN119757861A

  • A method for measuring the rotational frequency of magnetic particles based on wavelet transform

    CN119757861B

  • Data processing method and system for light-flow complementary power supply device in irrigated area

    CN120316415A