Shafting torsional vibration signal early warning method based on multi-scale morphological algorithm

The multi-scale morphological algorithm is used to preprocess and feature extraction of the torsional vibration signals of the shaft system, which solves the problems of insufficient signal analysis and early warning mechanism in the prior art, and realizes efficient analysis and early warning of the torsional vibration signals of the shaft system.

CN119086059APending Publication Date: 2024-12-06尹丽萍
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411254945.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-09
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The prior art lacks effective signal preprocessing and feature enhancement methods in the analysis of the torsional vibration signal of the shaft system, and lacks an adaptive threshold early warning mechanism, making it difficult to effectively deal with complex signal changes in the non-steady state speed change.

Method used

Using a method based on multi-scale morphological algorithm, the axis system torsional vibration signals are preprocessed, feature decomposition, morphological processing and modal analysis are carried out, and noise is removed through wavelet decomposition, morphological open operation and closed operation are eliminated to construct the axis system torsional vibrations. Modal function, and early warning signals are generated through feature extraction and comparison of multi-parameter and multi-threshold.

Benefits of technology

It significantly improves signal quality, enhances the sensitivity and accuracy of abnormal detection, realizes a flexible early warning mechanism, can timely and accurately reflect changes in the torsional vibration state of the shaft system, and improves the safe and stable operation of the equipment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119086059A_ABST
    Figure CN119086059A_ABST
Patent Text Reader

Abstract

The invention provides a shafting torsional vibration signal early warning method based on a multi-scale morphological algorithm, and the method comprises the steps: carrying out the preprocessing of an original shafting torsional vibration signal collected by a sensor, so as to obtain a shafting torsional vibration time sequence signal; performing characteristic decomposition on the shafting torsional vibration time sequence signal to obtain an abnormal shafting torsional vibration signal; performing morphological processing on the abnormal shaft system torsional vibration signal to obtain a noise-reduced shaft system torsional vibration signal; performing modal processing on the noise-reduced shaft system torsional vibration signal to construct a shaft system torsional vibration modal function; and carrying out feature extraction on the modal function to obtain a shaft system torsional vibration feature, and carrying out feature comparison with a set modal threshold to generate an early warning signal. In conclusion, according to the shafting torsional vibration signal early warning method based on the multi-scale morphological algorithm, a signal preprocessing and enhancing means is provided, a set of complete early warning mechanism is designed, and a solution is provided for shafting torque signal processing in the unstable speed change process.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of torsional vibration signal processing, and in particular relates to a shaft system torsional vibration signal early warning method based on a multi-scale morphological algorithm. Background Art

[0002] As a key component of rotating mechanical equipment, the shaft system is responsible for the transmission of torque. Its stable operation is crucial to ensure the safety and efficiency of mechanical equipment. However, the existence of torsional vibration (torsional vibration) poses a severe challenge to the health of the shaft system and the long-term operation of the equipment. Torsional vibration is latent, sudden and destructive. If it is not effectively controlled, it may lead to increased noise, accelerated wear of parts, and even cause shaft system fracture, resulting in significant economic losses and safety hazards. In order to meet this challenge, the accurate measurement and in-depth analysis of shaft system torsional vibration has become a key link to ensure the safe operation of mechanical equipment. Although the existing torsional vibration testing technology has made significant progress in signal extraction and transmission, it still faces significant technical challenges in the core link of signal analysis. This is mainly due to the complexity of the torsional vibration signal, especially the interweaving of multiple loads (such as torque, bending moment, axial force and unbalanced inertial force), which makes signal analysis extremely difficult.

[0003] In view of the above problems, in the existing technology, "Method for Identifying Torsional Vibration Modal Parameters of Pumped Storage Unit Shaft System Based on Strain Testing Technology" discloses a technical solution of arranging torque strain gauges on the exposed shaft section of the pumped storage unit and collecting torque pulsation data during load rejection through wireless devices. In another patent document "Method for Extracting Instantaneous Frequency Features of Torsional Vibration Signals Based on Improved Whale Optimization VMD", a method for extracting instantaneous frequency features of torsional vibration signals based on improved whale optimization VMD is disclosed. The main process of this method is to use energy entropy as the fitness function of the improved whale optimization algorithm, and use the improved whale optimization algorithm to accurately extract the parameters of the VMD algorithm; then substitute the extracted optimal parameters into the VMD algorithm to realize the noise reduction decomposition of the torsional vibration signal, and then use the intrinsic mode function IMF with a large correlation coefficient to realize the reconstruction of the torsional vibration signal, and then perform zero-point interpolation on the reconstructed torsional vibration signal, calculate the interval between the two pulses, calculate the time difference, calculate the instantaneous speed fluctuation of the torsional vibration, and finally perform spectrum analysis to realize the extraction of the instantaneous frequency features of the torsional vibration signal.

[0004] The above two patents provide a method for torsional vibration signal processing, but there are still the following deficiencies: the existing technology mainly focuses on the collection of torque pulsation data and the identification of modal parameters, but does not explicitly mention the technical means of signal preprocessing and feature enhancement. Moreover, the existing technology lacks an adaptive threshold warning mechanism. In addition, the processing of non-steady-state speed change processes is limited: although the existing technology aims to solve the problem that torsional vibration signals are difficult to demodulate during non-steady-state speed change processes, it does not describe in detail how to deal with complex signal changes during non-steady-state speed change processes. Summary of the invention

[0005] Based on the above problems, an embodiment of the present application provides a shaft system torsional vibration signal warning method based on a multi-scale morphological algorithm to solve the above problems existing in the prior art.

[0006] The embodiment of the present application discloses a shaft torsional vibration signal warning method based on a multi-scale morphological algorithm, comprising: preprocessing the original shaft torsional vibration signal collected by the sensor to obtain a shaft torsional vibration time series signal; performing feature decomposition on the shaft torsional vibration time series signal to obtain an abnormal shaft torsional vibration signal; performing morphological processing on the abnormal shaft torsional vibration signal to obtain a noise-reduced shaft torsional vibration signal; performing modal processing on the noise-reduced shaft torsional vibration signal to construct a shaft torsional vibration modal function; performing feature extraction on the modal function to obtain a shaft torsional vibration feature, and performing feature comparison with a set modal threshold to generate a warning signal.

[0007] Optionally, the characteristic decomposition of the shaft system torsional vibration time series signal to obtain the abnormal shaft system torsional vibration signal includes: performing wavelet decomposition on the shaft system torsional vibration time series signal to obtain approximate signals and detail signals at different scales; and performing statistical characteristic analysis on the approximate signals and the detail signals to obtain the abnormal shaft system torsional vibration signal.

[0008] Optionally, based on the following formula, wavelet decomposition is performed on the shaft system torsional vibration time series signal to obtain approximate signals and detail signals at different scales: Among them, c j,k is the wavelet coefficient of the approximate signal, d j,k is the wavelet coefficient of the detail signal, φ j,k (n) is the scale function, ψ j,k (n) is the wavelet function, j is the scale parameter, k is the translation parameter, J is the summation termination value, and n is the discrete time index.

[0009] Optionally, calculate the scaling function based on the following formula: j,k (n) = 2 -j / 2 φ(2 -j nk); where φ j,k (n) is the scaling function, n is the discrete time index, k is the translation parameter, and j is the scaling parameter.

[0010] Optionally, the wavelet function is calculated based on the following formula: j,k (n) = 2 -j / 2 ψ(2 -j nk) where ψ j,k (n) is the wavelet function, n is the discrete time index, k is the translation parameter, and j is the scale parameter.

[0011] Optionally, the abnormal shaft system torsional vibration signal is morphologically processed to obtain a noise-reduced shaft system torsional vibration signal, including: performing a morphological opening operation on the peak noise in the approximate signal and the detail signal to remove the peak noise and obtain a de-peaked signal; performing a morphological closing operation on the concave noise in the approximate signal and the detail signal to remove the concave noise and obtain a de-valleyed signal; and obtaining a noise-reduced shaft system torsional vibration signal based on the de-peaked signal and the de-valleyed signal.

[0012] Optionally, the modal processing of the noise-reduced shaft system torsional vibration signal to construct a shaft system torsional vibration modal function includes: reconstructing the modal of the noise-reduced shaft system torsional vibration signal to obtain a modal function; and screening characteristics of the modal function to construct a shaft system torsional vibration modal function based on the modal function that meets preset conditions.

[0013] Optionally, reconstructing the mode of the noise reduction shaft system torsional vibration signal to obtain a modal function includes: reconstructing the mode of the noise reduction shaft system torsional vibration signal based on a morphological opening operation to obtain a morphological opening operation function; reconstructing the mode of the noise reduction shaft system torsional vibration signal based on a morphological closing operation to obtain a morphological closing operation function; and obtaining the modal function based on the morphological opening operation function and the morphological closing operation function.

[0014] Optionally, the characteristics of the modal function are screened to construct a shaft system torsional vibration modal function based on the modal function that meets preset conditions, including: setting a frequency range based on the instantaneous frequency of the modal function; calculating the statistical characteristics of the instantaneous frequency; setting a screening criterion based on the frequency range and the statistical characteristics; selecting the modal function that meets the conditions according to the screening criterion to construct the shaft system torsional vibration modal function;

[0015] Optionally, feature extraction is performed on the modal function to obtain the torsional vibration feature of the shaft system, and feature comparison is performed with the set modal threshold to generate a warning signal, including: feature extraction is performed on the modal function to obtain feature information, and the feature information includes: instantaneous frequency, modal average instantaneous frequency, modal energy, modal peak, and modal kurtosis; based on the feature information, feature comparison is performed with the set modal threshold, and the modal threshold includes: average instantaneous frequency threshold, energy threshold, peak threshold, and kurtosis threshold; if at least any one of the following conditions is met, a warning signal is generated: the modal average instantaneous frequency is greater than the average instantaneous frequency threshold; the modal energy is equal to the energy threshold; the modal peak is equal to the peak threshold; the modal kurtosis is equal to the kurtosis threshold.

[0016] The present invention provides a shaft torsional vibration signal warning method based on a multi-scale morphological algorithm, including: pre-processing the original shaft torsional vibration signal collected by the sensor to obtain a shaft torsional vibration time series signal; feature decomposing the shaft torsional vibration time series signal to obtain an abnormal shaft torsional vibration signal; morphologically processing the abnormal shaft torsional vibration signal to obtain a de-noised shaft torsional vibration signal; modal processing the de-noised shaft torsional vibration signal to construct a shaft torsional vibration modal function; feature extraction of the modal function to obtain shaft torsional vibration features, and feature comparison with a set modal threshold to generate a warning signal. In the present application, the peak noise and concave noise in the shaft torsional vibration signal are effectively removed by morphological processing (such as morphological opening and closing operations) to obtain a de-noised shaft torsional vibration signal. The morphological method performs well in processing nonlinear signals and removing noise, can significantly improve signal quality, and provide a more accurate data basis for subsequent analysis. In addition, by using wavelet decomposition technology, this method can perform multi-scale decomposition on the shaft torsional vibration time series signal to obtain approximate signals and detail signals at different scales. This multi-scale analysis can reveal the characteristics of the signal in different frequency bands, help to more comprehensively understand the complex structure of the signal, and improve the sensitivity and accuracy of anomaly detection. In addition, by performing statistical characteristic analysis (such as mean, standard deviation, skewness, kurtosis, etc.) on the approximate signal and detail signal, combined with the set abnormal threshold, this method can accurately identify the abnormal situation in the shaft torsional vibration signal. This method not only considers the overall trend of the signal, but also pays attention to the local details of the signal, and realizes the comprehensive capture of abnormal signals. Finally, after feature extraction of the modal function, this method realizes a flexible early warning mechanism by setting multiple modal thresholds (such as average instantaneous frequency threshold, energy threshold, peak threshold, kurtosis threshold) and comparing them with the extracted feature information. This multi-parameter, multi-threshold early warning method improves the reliability and pertinence of the early warning signal, can timely and accurately reflect the changes in the shaft torsional vibration state, and provides strong support for equipment maintenance and troubleshooting. In summary, the shaft torsional vibration signal warning method based on a multi-scale morphological algorithm provided in this application provides a means of signal preprocessing and enhancement, designs a complete warning mechanism, and provides a solution for shaft torque signal processing during non-steady-state speed change. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0018] Figure 1This is a flow chart of a shaft system torsional vibration signal warning method based on a multi-scale morphological algorithm in Example 1 of the present application. DETAILED DESCRIPTION

[0019] The implementation of any technical solution of the embodiments of the present application does not necessarily require achieving all of the above advantages at the same time.

[0020] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Figure 1 Flow chart of a shaft torsional vibration signal warning method based on a multi-scale morphological algorithm in Example 1 of the present application; Figure 1As shown, a shaft torsional vibration signal warning method based on a multi-scale morphological algorithm includes: pre-processing the original shaft torsional vibration signal collected by the sensor to obtain a shaft torsional vibration time series signal; performing feature decomposition on the shaft torsional vibration time series signal to obtain an abnormal shaft torsional vibration signal; performing morphological processing on the abnormal shaft torsional vibration signal to obtain a de-noised shaft torsional vibration signal; performing modal processing on the de-noised shaft torsional vibration signal to construct a shaft torsional vibration modal function; performing feature extraction on the modal function to obtain shaft torsional vibration features, and performing feature comparison with a set modal threshold to generate a warning signal. In this embodiment, the peak noise and concave noise in the shaft torsional vibration signal are effectively removed by morphological processing (such as morphological opening and closing operations) to obtain a de-noised shaft torsional vibration signal. The morphological method performs well in processing nonlinear signals and removing noise, and can significantly improve signal quality, providing a more accurate data basis for subsequent analysis. In addition, by using wavelet decomposition technology, this method can perform multi-scale decomposition on the shaft torsional vibration time series signal to obtain approximate signals and detail signals at different scales. This multi-scale analysis can reveal the characteristics of the signal in different frequency bands, help to more comprehensively understand the complex structure of the signal, and improve the sensitivity and accuracy of anomaly detection. In addition, by performing statistical characteristic analysis (such as mean, standard deviation, skewness, kurtosis, etc.) on the approximate signal and detail signal, combined with the set abnormal threshold, this method can accurately identify the abnormal situation in the shaft torsional vibration signal. This method not only considers the overall trend of the signal, but also pays attention to the local details of the signal, and realizes the comprehensive capture of abnormal signals. Finally, after feature extraction of the modal function, this method realizes a flexible early warning mechanism by setting multiple modal thresholds (such as average instantaneous frequency threshold, energy threshold, peak threshold, kurtosis threshold) and comparing them with the extracted feature information. This multi-parameter, multi-threshold early warning method improves the reliability and pertinence of the early warning signal, can timely and accurately reflect the changes in the shaft torsional vibration state, and provides strong support for equipment maintenance and troubleshooting. In summary, the shaft torsional vibration signal warning method based on a multi-scale morphological algorithm provided in this application can improve the accuracy and robustness of signal analysis, thereby more effectively monitoring and controlling the torsional vibration of the shaft system and ensuring the safe and stable operation of mechanical equipment.

[0022] Optionally, the preprocessing includes: removing DC offset and filtering high-frequency noise. In the present embodiment, DC offset is usually introduced by the deviation of the sensor or signal processing system, which will cause the overall level of the signal to deviate from the baseline. By removing the DC offset, the baseline of the signal can be regressed to zero or a set reference point, thereby improving the overall quality of the signal and providing a more accurate data basis for subsequent analysis and processing. In addition, high-frequency noise is usually caused by environmental interference, sensor noise or interference during signal transmission. These noises will be superimposed on useful signals, reduce the signal-to-noise ratio of the signal, and affect the resolution and accuracy of the signal. By filtering out high-frequency noise, the interference of noise on signal analysis can be reduced, and the clarity and credibility of the signal can be improved. In addition, removing DC offset and filtering out high-frequency noise in the preprocessing step can highlight useful features in the signal, such as low-frequency trends, periodic components, etc. These features are crucial for subsequent signal analysis, feature extraction and pattern recognition. Through preprocessing, these features can be made more obvious and easy to be identified and extracted.

[0023] Optionally, the characteristic decomposition of the shaft system torsional vibration time series signal to obtain the abnormal shaft system torsional vibration signal includes: performing wavelet decomposition on the shaft system torsional vibration time series signal to obtain approximate signals and detail signals at different scales; and performing statistical characteristic analysis on the approximate signal and the detail signal to obtain the abnormal shaft system torsional vibration signal. In this embodiment, the shaft system torsional vibration time series signal is decomposed into approximate signals and detail signals at different scales by wavelet decomposition technology. This multi-scale analysis method can simultaneously capture the characteristics of the signal in the low-frequency and high-frequency ranges, making the analysis of the signal more comprehensive and detailed. By performing statistical characteristic analysis (such as mean, standard deviation, skewness, kurtosis, etc.) on the approximate signal and the detail signal, the abnormal components in the signal, such as peak noise and concave noise, can be effectively identified and extracted. This statistical characteristic analysis provides a scientific basis for subsequent noise reduction processing and helps to improve the accuracy and efficiency of signal processing. Moreover, the approximate signal and detail signal obtained by wavelet decomposition can be further subjected to morphological processing, such as morphological opening and closing operations, to remove the noise components in the signal and retain useful information. This method helps to extract the features of the shaft torsional vibration signal more accurately, and provides a high-quality data basis for subsequent feature extraction and pattern recognition. In addition, by performing modal processing on the shaft torsional vibration signal after noise reduction, constructing the shaft torsional vibration modal function, and performing feature extraction and modal threshold comparison based on the modal function, a flexible early warning mechanism can be implemented. This early warning mechanism can adjust the early warning threshold according to the actual situation of the signal, thereby improving the accuracy and reliability of the early warning. Finally, this embodiment combines multiple technical means such as multi-scale analysis, statistical characteristic analysis, morphological processing and modal analysis, so that the entire early warning system has strong adaptability and robustness. Whether it is for different types of shaft torsional vibration signals or for different types of noise in the signal, this method can effectively process and analyze.

[0024] The approximate signal is the overall trend or slow change of the signal in the low-frequency range. In wavelet transform, the approximate signal is obtained through a low-pass filter (LF), which reflects the smoother part of the signal, which can be a long-term trend, a low-frequency component, etc. The approximate signal is of great significance for analyzing the overall trend, low-frequency components, or removing high-frequency noise of the signal. By continuously decomposing the approximate signal, the signal can be decomposed into multiple low-frequency components, which are very useful for applications such as signal compression, denoising, or feature extraction.

[0025] The detail signal is the detail information of the signal in the high frequency range. In wavelet transform, the detail signal is obtained through a high-pass filter (HF), which reflects the rapidly changing part of the signal, which can be a spike, a sharp edge, etc. The detail signal is very useful for analyzing the local characteristics, mutation points or high-frequency noise of the signal. Through multi-level wavelet decomposition, detail signals at different scales (or different frequency ranges) can be obtained, so as to understand the high-frequency characteristics of the signal in more detail.

[0026] Optionally, based on the following formula, wavelet decomposition is performed on the shaft system torsional vibration time series signal to obtain approximate signals and detail signals at different scales: Among them, c j,k is the wavelet coefficient of the approximate signal, d j,k is the wavelet coefficient of the detail signal, φ j,k (n) is the scale function, ψ j,k (n) is the wavelet function, j is the scale parameter, k is the translation parameter, J is the summation termination value, and n is the discrete time index. In this embodiment, wavelet decomposition can decompose the signal into approximate signals and detail signals at different scales (or different frequency ranges). This multi-scale analysis capability enables the method to more comprehensively capture the low-frequency trends and high-frequency details in the shaft system torsional vibration signal, thereby more accurately reflecting the operating status of the shaft system. Moreover, through wavelet decomposition, the high-frequency noise is mainly concentrated in the detail signal, while the low-frequency trend is retained in the approximate signal. This characteristic enables the method to effectively suppress high-frequency noise, improve the signal-to-noise ratio of the signal, and provide a clearer data basis for subsequent signal processing and analysis. In addition, the decomposed approximate signal and detail signal represent the low-frequency trend and high-frequency details of the signal, respectively, which enables the key features of the shaft system torsional vibration signal, such as instantaneous frequency, modal energy, etc., to be more accurately extracted in the subsequent feature extraction process, thereby improving the accuracy and reliability of the early warning. Finally, the wavelet decomposition method is highly flexible and scalable. By selecting different scale functions and wavelet functions, customized analysis can be performed for different types of shaft system torsional vibration signals. In addition, this method can also be combined with other signal processing technologies to form a more complex and powerful signal processing system to meet the needs of different application scenarios.

[0027] Optionally, based on the scale function, the shaft system torsional vibration time series signal is subjected to wavelet decomposition to obtain approximate signals at different scales. In this embodiment, by using the scale function for wavelet decomposition, the signal can be decomposed into approximate signals and detail signals of multiple scales. This multi-scale analysis can simultaneously capture the characteristics of the signal at different frequencies or time scales, thereby providing a more comprehensive understanding of the signal. Moreover, the approximate signal usually reflects the overall trend or slow change of the signal in the low-frequency range. This part of the signal is usually smoother and contains the main features of the signal while avoiding the interference of high-frequency noise. Therefore, by decomposing and focusing on the approximate signal, high-frequency noise can be more effectively removed and the signal-to-noise ratio of the signal can be improved. In addition, the scale function of the wavelet decomposition can be selected according to the specific application requirements, thereby achieving flexibility and personalized customization of signal processing. This flexibility enables the method to adapt to different processing requirements for signals in different scenarios. Finally, by performing detailed analysis on the approximate signals at different scales, the characteristic information of the signal can be extracted more accurately. These characteristic information are crucial for subsequent applications such as abnormality detection and fault diagnosis.

[0028] Optionally, calculate the scaling function based on the following formula: j,k (n) = 2 -j / 2 φ(2 -j nk); where φ j,k (n) is the scale function, n is the discrete time index, and k is the translation parameter, which indicates the position of the wavelet function on the time axis. Different k values ​​mean that the wavelet function is matched at different positions of the signal. j is the scale parameter, a larger j indicates a larger scale (lower resolution), and a smaller j indicates a smaller scale (higher resolution). In this embodiment, by providing a specific mathematical formula, this method ensures the clarity and accuracy of the wavelet decomposition. Developers can directly apply this formula to calculate the scale function without additional approximations or assumptions, thereby improving the reliability and accuracy of the entire signal processing process. Moreover, the variables in the formula (such as the number of sampling points n and the frequency index k) allow flexible adjustment of different signal characteristics, and can be optimized according to the specific characteristics of the shaft system torsional vibration signal to adapt to different working conditions and system requirements.

[0029] Optionally, the shaft system torsional vibration signal warning method based on the multi-scale morphological algorithm performs wavelet decomposition on the shaft system torsional vibration time series signal based on the scale function, and obtains approximate signals at different scales according to the following formula: Where n is the discrete time index and k is the translation parameter. In this embodiment, through wavelet decomposition, the method can analyze the shaft torsional vibration signal at different scales (or frequency ranges). This multi-scale analysis capability can more comprehensively understand the dynamic characteristics of the signal, including the overall trend of low frequencies and the local details of high frequencies. Moreover, the use of scale functions for wavelet decomposition can accurately capture the approximate components of the signal at different scales, which can better retain the time-frequency characteristics of the signal than simple frequency domain analysis methods, thereby improving the accuracy of the analysis. In addition, by adjusting the parameters in the scale function (such as n and k in the formula), the method can be flexibly configured according to the characteristics of the actual signal and the analysis requirements, thereby improving the adaptability and universality of the method.

[0030] Optionally, based on the wavelet function, the shaft system torsional vibration time series signal is subjected to wavelet decomposition to obtain approximate signals at different scales. In this embodiment, wavelet decomposition can decompose the signal into approximate signals and detail signals at different scales (or different frequency ranges). This allows analysts to understand the characteristics of the signal in different frequency bands more carefully, so as to capture the dynamic behavior of the shaft system torsional vibration signal more comprehensively. Moreover, through wavelet decomposition, the high-frequency part of the signal (i.e., the detail signal) can be separated from the low-frequency part (i.e., the approximate signal). Since high-frequency noise is usually contained in the detail signal, the noise can be suppressed by processing or ignoring the detail signal to improve the signal-to-noise ratio. In addition, the detail signal reflects the rapidly changing parts of the signal, such as spikes, sharp edges, etc., which is very useful for extracting local features of the signal. In the analysis of shaft system torsional vibration signals, these local features may indicate faults or abnormal conditions, so accurate extraction of these features is crucial for fault diagnosis and early warning. Finally, compared with the traditional Fourier transform, the wavelet transform has a stronger ability in processing non-stationary signals. The shaft torsional vibration signal is often non-stationary, containing multiple frequency components and transient changes, so wavelet transform is more suitable for the analysis of shaft torsional vibration signals.

[0031] Optionally, the wavelet function is calculated based on the following formula: j,k (n) = 2 - / 2ψ(2 -j nk) where ψ j,k(n) is a wavelet function, n is a discrete time index, k is a translation parameter, and j is a scale parameter. In this embodiment, through wavelet transform, approximate signals and detail signals of the signal at different scales (or different frequency ranges) can be obtained. This multi-scale analysis capability enables the method to capture the overall trend and local details of the signal at the same time, providing rich information for subsequent anomaly detection and feature extraction. In addition, wavelet decomposition obtains approximate signals through a low-pass filter (LF), which usually reflect the smoother parts of the signal, such as long-term trends and low-frequency components, thereby effectively suppressing high-frequency noise. This is very helpful for improving the signal-to-noise ratio of the signal and the accuracy of subsequent processing. At the same time, the detail signal obtained by the high-pass filter (HF) contains the rapidly changing parts of the signal, such as spikes and sharp edges, which is very useful for analyzing the local features and mutation points of the signal. These detail information plays a key role in anomaly detection and feature recognition.

[0032] Optionally, based on the wavelet function, the shaft system torsional vibration time series signal is subjected to wavelet decomposition, and detail signals at different scales are obtained according to the following formula: Among them, d j (n) is the detail signal, n is the discrete time index, k is the translation parameter, and j is the scale parameter. In this embodiment, wavelet decomposition can provide detailed information of the signal at different scales (or frequency ranges), which enables analysts to understand the high-frequency characteristics of the signal in more detail. For the shaft torsional vibration signal, information at different scales may reflect the characteristics of different physical processes, such as mechanical vibration, imbalance of the transmission system, etc. Moreover, through multi-level wavelet decomposition, the high-frequency noise in the signal can be separated from the useful signal. Detail signals often contain high-frequency components in the signal, including noise and mutation points, and through the analysis and processing of these signals, the noise can be effectively suppressed or removed, and the signal-to-noise ratio of the signal can be improved. Finally, the shaft torsional vibration signal early warning method based on the multi-scale morphological algorithm can more accurately identify the abnormal features in the shaft torsional vibration signal through in-depth analysis of the detail signal, thereby improving the accuracy of the early warning.

[0033] Optionally, the statistical characteristic analysis of the approximate signal and the detail signal to obtain the abnormal shaft system torsional vibration signal includes: calculating the mean of the approximate signal and the detail signal to obtain the central trend of the shaft system torsional vibration signal; calculating the standard deviation of the approximate signal and the detail signal to obtain the degree of fluctuation of the system torsional vibration signal; calculating the skewness of the approximate signal and the detail signal to obtain the distribution asymmetry of the system torsional vibration signal; calculating the kurtosis of the approximate signal and the detail signal to obtain the sharpness of the system torsional vibration signal; based on the central trend, the degree of fluctuation, the distribution asymmetry and the sharpness, the abnormal shaft system torsional vibration signal is obtained. In this embodiment, by calculating statistical quantities such as the mean, standard deviation, skewness and kurtosis, the embodiment can comprehensively analyze the characteristics of the shaft system torsional vibration signal from multiple dimensions. This can not only reflect the central trend and fluctuation degree of the signal, but also capture the asymmetry and sharpness of its distribution, providing a more comprehensive basis for identifying abnormal signals. In addition, these statistics can be used to accurately identify abnormal parts in the shaft torsional vibration signal, such as spike noise and concave noise. This multi-dimensional analysis method greatly improves the accuracy and reliability of abnormal signal identification compared to single-dimensional analysis. Moreover, by setting a threshold and comparing it with the calculated statistics, it can respond to slight changes in the shaft torsional vibration signal, with high sensitivity. At the same time, due to the integration of information from multiple statistics, it also has a certain degree of robustness and can resist the influence of noise and other interference factors to a certain extent. Finally, through the timely discovery and identification of abnormal signals, early warning can be provided before the shaft torsional vibration problem further develops into a serious fault. This helps operators take timely measures to avoid or reduce equipment damage and downtime, and reduce maintenance costs.

[0034] Optionally, the mean value of the approximate signal is calculated based on the following formula: Among them, μ j,c is the mean of the approximate signal, c j(i) is an approximate signal, and N is the number of approximate signals. In this embodiment, by calculating the mean of the approximate signal, the central trend or overall stability level of the shaft system torsional vibration signal in the low-frequency range can be quantitatively evaluated, which helps to more accurately understand the behavioral characteristics of the signal in a specific time period, thereby making it easier to identify abnormal phenomena that deviate from the normal state. Moreover, as a basic statistical characteristic of the signal, the mean plays an important role in identifying abnormal changes in the signal. By comparing the mean of the actual signal with the set threshold or historical data, it is possible to more effectively determine whether the shaft system torsional vibration signal is in a normal state, thereby improving the accuracy of the early warning system. In addition, the mean is not only a measure of the central trend of the signal, but also an important input for subsequent signal processing steps. In the subsequent processes of morphological processing, modal function construction, and feature extraction, the calculation result of the mean will be used as a reference or benchmark, which will help to further optimize the performance of the entire early warning method.

[0035] Optionally, the mean value of the detail signal is calculated based on the following formula: Among them, μ j,d is the mean value of the detail signal, d j (i) is the detail signal, N is the number of detail signals, μ j,c is the mean of the approximate signal. In this embodiment, by associating the mean of the detail signal with the mean of the approximate signal, the central trend of the entire shaft system torsional vibration signal can be more effectively evaluated. In addition, the detail signal usually contains high-frequency noise and rapidly changing information, and calculating its mean alone may be greatly affected by the noise. By combining it with the mean of the approximate signal, the influence of high-frequency noise can be partially offset, thereby reducing the error. Moreover, in multi-scale morphological processing, the approximate signal and detail signal obtained by wavelet decomposition of the signal represent the low-frequency and high-frequency parts of the signal respectively in a physical sense. Combining the two to calculate the mean can make the algorithm more robust to different types of shaft system torsional vibration signals and adapt to a wider range of working conditions and noise environments. Finally, in the subsequent feature extraction and early warning judgment process, the mean based on this comprehensive calculation can more accurately reflect the characteristics of the shaft system torsional vibration signal, thereby optimizing the effect of feature extraction and improving the accuracy and reliability of the early warning system.

[0036] Optionally, the standard deviation of the approximate signal is calculated based on the following formula: Among them, σ j,c is the standard deviation of the approximate signal, c j (i) is the approximate signal, μ j,cis the mean of the approximate signal, and N is the number of approximate signals. In this embodiment, when calculating the mean of the detail signal, the mean of the approximate signal is considered for correction, so that the central trend of the shaft system torsional vibration signal in the high-frequency detail part can be more accurately reflected. Since the detail signal itself reflects the high-frequency rapidly changing part, it is easily affected by noise, and the mean of the approximate signal provides a relatively stable benchmark, which helps to reduce errors. In addition, by introducing the mean of the approximate signal as a correction term, the influence of high-frequency noise on the mean calculation of the detail signal can be suppressed to a certain extent. This method is of great significance for improving the robustness of signal processing and reducing misjudgment. In addition, considering the mean of the approximate signal and the detail signal at the same time makes the analysis of the shaft system torsional vibration signal more comprehensive. The approximate signal reflects the overall trend of the signal, while the detail signal reflects the local details of the signal. The combination of the two can more completely understand the characteristics of the signal. Finally, this processing method enhances the applicability of the early warning method under different working conditions and noise environments. By introducing the correction term, the method can still maintain high accuracy and stability when facing shaft system torsional vibration signals of different complexities.

[0037] Optionally, the standard deviation of the detail signal is calculated based on the following formula: Among them, σ j,c is the standard deviation of the detail signal, d j (i) is the detail signal, μ j,d is the mean of the detail signal, and N is the number of detail signals. In this embodiment, the standard deviation is an important indicator to measure the degree of data discreteness. By calculating the standard deviation of the detail signal, the fluctuation or change degree of the shaft system torsional vibration signal in the high frequency range can be accurately evaluated. Moreover, the detail signal represents the detailed information of the signal in the high frequency range, which is obtained through a high-pass filter and reflects the rapidly changing part of the signal. Calculating its standard deviation helps to understand the high-frequency characteristics of the signal in more detail and provide an important basis for subsequent fault diagnosis or early warning. In addition, since the standard deviation can reflect the degree of data discreteness, when there are abnormal rapid fluctuations in the shaft system torsional vibration signal, the standard deviation of these detail signals will increase significantly. Therefore, by monitoring the standard deviation of the detail signal, the sensitivity of abnormal detection can be improved and potential problems can be discovered in time. Finally, compared with statistical characteristics such as mean, skewness, and kurtosis, the standard deviation provides an independent perspective on the degree of signal fluctuation. Combining these statistical characteristics for analysis can more comprehensively evaluate the state of the shaft system torsional vibration signal and improve the accuracy and reliability of the early warning system.

[0038] Optionally, the skewness of the approximate signal is calculated based on the following formula: Among them, σ j,c is the standard deviation of the approximate signal, c j (i) is the approximate signal, μ j,cis the mean of the approximate signal, and N is the number of approximate signals. In this embodiment, the skewness is a statistic that measures the direction and degree of skewness of the data distribution. In the analysis of the shaft torsional vibration signal, the skewness can help us quantitatively analyze whether the signal waveform is symmetrical, so as to determine whether there are abnormal asymmetric characteristics in the signal, such as spike noise or concave noise. In addition, by calculating the skewness, the abnormal part in the shaft torsional vibration signal can be more accurately identified and processed, thereby avoiding the introduction of additional errors in the signal analysis process. This is crucial to ensure accurate monitoring and early warning of the shaft operation status. In addition, in the multi-scale morphological algorithm, by calculating the skewness of the approximate signal at different scales, the characteristic changes of the signal in different frequency ranges can be further analyzed, providing strong support for a comprehensive understanding of the shaft torsional vibration signal. Finally, the formulated calculation method ensures the standardization and repeatability of the results, making the analysis results at different times and under different conditions comparable, and improving the reliability and practicality of the analysis results.

[0039] Optionally, the skewness of the detail signal is calculated based on the following formula: Among them, γ j,d is the distribution asymmetry, σ j,d is the standard deviation of the detail signal, d j (i) is the detail signal, μ j,d is the mean of the detail signal, and N is the number of detail signals. In this embodiment, the distribution asymmetry provides an in-depth understanding of the characteristics of the data set, which helps analysts identify outliers, extreme values, or atypical situations in the data.

[0040] Optionally, the kurtosis of the approximate signal is calculated based on the following formula: Among them, β j,c is the kurtosis of the approximate signal, c j (i) is the approximate signal, μ j,cis the mean of the approximate signal, and N is the number of approximate signals. In this embodiment, kurtosis is a statistic that describes the sharpness of the signal distribution. By calculating the kurtosis of the approximate signal, the sharpness of the shaft torsional vibration signal can be accurately quantified. This is very useful for identifying extreme changes or abnormal conditions in the signal. Moreover, signals with high kurtosis usually contain more extreme values ​​or peaks, which may be manifestations of abnormal shaft torsional vibration. By monitoring the changes in kurtosis, abnormal conditions in shaft torsional vibration can be detected more sensitively, so as to provide early warning or intervention. In addition, kurtosis can be used as a dimension for analyzing shaft torsional vibration signals, and can be combined with other statistical characteristics (such as mean, standard deviation, skewness, etc.) to comprehensively evaluate the state of shaft torsional vibration from multiple angles. This multi-dimensional analysis method can improve the accuracy and reliability of fault diagnosis. Finally, the calculation method of kurtosis does not depend on the specific form or structure of the signal, so it has good adaptability. Regardless of the type of shaft torsional vibration signal, as long as enough sample data can be obtained, its kurtosis can be calculated, and then abnormal detection and diagnosis can be performed.

[0041] Optionally, the kurtosis of the detail signal is calculated based on the following formula: Among them, β j,d is the kurtosis of the detail signal, d j (i) is the detail signal, μ j,d is the mean of the detail signal, and N is the number of detail signals. In this embodiment, by calculating the kurtosis of the detail signal, the sharpness or local characteristics of the shaft system torsional vibration signal in the high frequency range can be further understood, which helps to capture the subtle changes of the signal more comprehensively, thereby more accurately identifying potential anomalies or faults. Moreover, kurtosis is a statistic that measures the sharpness of the data distribution. A high kurtosis indicates that there are many extreme values ​​far from the mean in the data, which may correspond to spike noise or mutation points in the signal. The kurtosis information in these detail signals, combined with other statistical characteristics (such as mean, standard deviation, skewness, etc.), can more effectively identify abnormal shaft system torsional vibration signals and improve the accuracy of the early warning system. In addition, by decomposing and analyzing the statistical characteristics of the signal at different scales through a multi-scale morphological algorithm, including the calculation of the kurtosis of the detail signal, the system can be made more robust to different types of noise and interference. Even if there are complex noise components in the signal, the system can accurately identify abnormal signals by integrating multiple statistical characteristics. Finally, kurtosis, as one of the signal features, can be used together with other features to dynamically adjust and optimize the parameter settings of the early warning system. For example, according to the changes in historical data and real-time signal characteristics, parameters such as the kurtosis threshold can be automatically adjusted to adapt to different working conditions and operating environments, thereby improving the system's adaptability and intelligence.

[0042] Optionally, the abnormal shafting torsional vibration signal is obtained based on the central trend, the fluctuation degree, the distribution asymmetry and the sharpness, including: calculating the abnormal threshold of the shafting torsional vibration signal, if the abnormal threshold is equal to the mean, then the shafting torsional vibration signal has peak noise, if the abnormal threshold is equal to the set empirical value, then the shafting torsional vibration signal has concave noise, so as to obtain the abnormal shafting torsional vibration signal through the peak noise and the concave noise. In this embodiment, by calculating the mean, standard deviation, skewness and kurtosis of the signal, the statistical characteristics of the signal can be fully considered, so as to more accurately identify the abnormal part of the signal. These statistical characteristics reflect the central trend, fluctuation degree, distribution asymmetry and sharpness of the signal, which are very important for capturing abnormal features in the signal. Moreover, accurate abnormality detection is the basis of the reliability of the early warning system. Through the method of this embodiment, the abnormal part of the shafting torsional vibration signal can be more effectively identified, and an early warning signal can be generated accordingly. This helps to timely discover and deal with potential faults and avoid accidents, thereby improving the reliability and safety of the entire system. In addition, this embodiment distinguishes between two different types of noise, namely, peak noise and concave noise, and determines the noise type by comparing the abnormal threshold with the mean value or setting an empirical value. This distinguishing ability helps to take more targeted noise reduction measures in subsequent signal processing, thereby improving the efficiency and effect of signal processing.

[0043] Optionally, the abnormal shaft torsional vibration signal is obtained based on the central trend, the fluctuation degree, the distribution asymmetry and the sharpness, and further includes: judging whether the shaft torsional vibration signal is abnormal based on the skewness; if the skewness is greater than 0, the shaft torsional vibration signal may have peak noise; if the skewness is less than 0, the shaft torsional vibration signal may have concave noise; and obtaining the abnormal shaft torsional vibration signal through the peak noise and the concave noise. In this embodiment, by calculating the skewness, the distribution asymmetry of the shaft torsional vibration signal can be quantified, so as to more accurately judge whether there is an abnormality in the signal. A skewness greater than 0 indicates that the signal distribution is skewed to the right and there may be peak noise; less than 0 indicates that it is skewed to the left and there may be concave noise. This quantitative analysis is more accurate than relying solely on visual judgment. Moreover, as one of the statistical characteristics, the skewness has a certain anti-interference ability to random noise and interference in the signal. Therefore, the judgment method based on skewness can enhance the robustness of the entire early warning system, so that it can work effectively under different working conditions and noise environments. In addition, by timely identifying the peak noise and concave noise in the signal, the abnormality of the shaft torsional vibration can be discovered earlier. This helps to take preventive measures before the failure occurs, avoid equipment damage and downtime accidents, and thus improve production efficiency and safety. Finally, after identifying the abnormal shaft torsional vibration signal, it can be further processed by morphological processing, modal reconstruction and other steps to obtain more accurate shaft torsional vibration feature information.

[0044] Optionally, the abnormal shafting torsional vibration signal is obtained based on the central trend, the fluctuation degree, the distribution asymmetry and the sharpness, and further includes: judging whether the shafting torsional vibration signal is abnormal based on the kurtosis; if the kurtosis is greater than the set value, the shafting torsional vibration signal may have peak noise; if the kurtosis is less than the set value, the shafting torsional vibration signal may have concave noise. The abnormal shafting torsional vibration signal is obtained by the peak noise and the concave noise. In this embodiment, the kurtosis is a statistic for measuring the sharpness of the signal, which can reflect the sharpness or flatness of the distribution in the signal. By comparing the kurtosis with the set value, it is possible to accurately judge whether the signal contains peak noise or concave noise, and then accurately identify the abnormal shafting torsional vibration signal, thereby improving the accuracy of the analysis. Moreover, the kurtosis is a clear mathematical quantity, which is easy to calculate and compare. By setting a reasonable kurtosis threshold, it can be conveniently applied to the shafting torsional vibration signal warning in different scenarios, thereby improving the flexibility and applicability of the system. In addition, in the multi-scale morphological algorithm, the kurtosis analysis can be combined with technologies such as wavelet decomposition to perform multi-scale analysis on the signal. In this way, the sharpness of the signal can be observed more carefully at different scales, so as to more accurately identify abnormal shaft torsional vibration signals. Finally, the abnormal judgment based on kurtosis can quickly identify the peak noise and concave noise in the signal, reducing the workload of subsequent processing. At the same time, through the preset threshold conditions, the early warning signal can be triggered immediately, which improves the efficiency and response speed of the early warning.

[0045] Optionally, the abnormal shaft torsional vibration signal is subjected to morphological processing to obtain a reduced-noise shaft torsional vibration signal, including: performing a morphological opening operation on the peak noise in the approximate signal and the detail signal to remove the peak noise and obtain a de-peaked signal; performing a morphological closing operation on the concave noise in the approximate signal and the detail signal to remove the concave noise and obtain a de-valleyed signal; and obtaining a reduced-noise shaft torsional vibration signal based on the de-peaked signal and the de-valleyed signal. In this embodiment, the peak noise in the approximate signal and the detail signal is processed by the morphological opening operation, which can effectively remove these high-frequency and sharp noise components, making the signal smoother and reducing misjudgment caused by noise interference. At the same time, the morphological closing operation is used to process the concave noise in the approximate signal and the detail signal, filling the low-lying part in the signal, and further improving the purity of the signal. In addition, while removing noise, the morphological processing can better retain the original features of the signal, such as the edge and shape of the signal. Finally, the quality of the shaft torsional vibration signal after noise reduction is significantly improved, reducing the interference of noise on signal analysis, making the subsequent feature extraction, modal function construction and warning signal generation more accurate and reliable.

[0046] Optionally, based on the following formula, a morphological opening operation is performed on the peak noise in the approximate signal to remove the peak noise and obtain a de-peaked signal: where c j * (n) is the approximate signal after peak removal, c j (n) is an approximate signal, and b(n) is a structural element. In this embodiment, the morphological opening operation can effectively remove the spike noise in the signal, because these spikes usually appear as short-term, high-amplitude mutation points in the signal. Through the morphological opening operation, these spike noises are smoothed out, thereby improving the signal-to-noise ratio of the signal. Moreover, compared with traditional denoising methods, the morphological opening operation can better retain the detailed characteristics of the signal while removing noise. This is because the morphological operation operates based on the geometric shape of the signal, rather than simply filtering in the frequency domain. In addition, the effect of the morphological opening operation can be flexibly controlled by adjusting the size and shape of the structural element. This means that the most suitable structural element can be selected according to the characteristics of the signal and the distribution of the noise to achieve the best denoising effect. Finally, the calculation process of the morphological opening operation is relatively simple and easy to implement, so it has good real-time performance.

[0047] Optionally, based on the following formula, a morphological opening operation is performed on the peak noise in the detail signal to remove the peak noise and obtain a de-peaked signal: in is the detail signal after peak removal, d j (n) is the detail signal, and b(n) is the structure element. In this embodiment, the structure element is opened and applied to the input signal c j (n) and d j (n) to find the part of the signal that can be completely covered by the structure element. The opening operation is usually represented by Θ, and the closing operation is usually represented by c j * (n), It is the approximate signal and detail signal after the spike noise is removed. By performing morphological opening operation on the spike noise in the detail signal, these sharp and abnormal signal parts, namely the spike noise, can be accurately removed. The morphological opening operation can effectively smooth the peak part of the signal and retain other important features of the signal, thereby improving the purity and signal-to-noise ratio of the signal. In addition, after removing the spike noise, the obtained de-peaked signal is closer to the true state of the shaft torsional vibration signal. This helps to extract and analyze the signal more accurately in the future, reduce the impact of noise on the analysis results, and improve the accuracy and reliability of the early warning method. In addition, the structural elements in the morphological opening operation can be selected and designed according to the actual situation to adapt to different types and degrees of spike noise. This makes the method very flexible and adaptable, and can be applied to a variety of complex shaft torsional vibration signal early warning scenarios. Finally, after removing the spike noise, the subsequent feature extraction and comparison process will be more efficient and accurate. This helps the early warning system to identify potential shaft torsional vibration problems more quickly and issue early warning signals in a timely manner, thereby improving the early warning efficiency and response speed.

[0048] Optionally, based on the following formula, a morphological closing operation is performed on the concave noise in the approximate signal to remove the concave noise and obtain a valley-free signal: Among them, c j # (n) is the approximate signal after valley removal, c j (n) is the approximate signal and b(n) is the structural element.

[0049] Based on the following formula, a morphological closing operation is performed on the concave noise in the detail signal to remove the concave noise and obtain a valley-free signal: where d j # (n) is the detail signal after valley removal, d j (m) is the detail signal, and b(n) is the structure element. In this embodiment, the structure element b(n) is closed on the input signal c j (n) and d j (n), the concave noise in the signal is removed by dilation operation, and then the basic shape of the signal is restored by erosion operation. j # (n), d j #(n) is the approximate signal and detail signal after the concave noise is removed. Morphological closing operation is a basic operation in mathematical morphology, through which the concave noise in the signal (i.e., the negative mutation or valley part in the signal) can be effectively removed. This is very important for the shaft torsional vibration signal, because the concave noise may be caused by system failure or abnormality. By removing these noises, the signal can be made purer and convenient for subsequent analysis. Moreover, while removing the noise, the morphological closing operation can better retain the original characteristics of the signal, especially the overall trend and shape of the signal. This is very beneficial for subsequent feature extraction and modal analysis, because the original information of the signal is retained and the state of the shaft torsional vibration can be more accurately reflected. In addition, by performing morphological closing operation on the concave noise in the approximate signal, a smoother and continuous signal, i.e., a valley-free signal, can be obtained. This improvement in signal quality helps to improve the accuracy of subsequent feature decomposition and modal processing, thereby improving the performance of the entire early warning system. Finally, since the morphological closing operation can effectively remove the interference noise in the signal, the shaft system torsional vibration modal function constructed based on the valley-removing signal will more accurately reflect the true state of the shaft system, help improve the accuracy of the early warning system, and enable the system to issue early warning signals more timely and accurately, thereby avoiding potential failures or accidents.

[0050] Optionally, the modal processing of the noise-reduced shaft system torsional vibration signal to construct a shaft system torsional vibration modal function includes: reconstructing the modal of the noise-reduced shaft system torsional vibration signal to obtain a modal function; screening the characteristics of the modal function to construct a shaft system torsional vibration modal function based on the modal function that meets the preset conditions. In this embodiment, the quality of the signal is further improved by modally reconstructing the noise-reduced shaft system torsional vibration signal. Modal reconstruction uses advanced technologies such as morphological opening and closing operations to effectively remove peak noise and concave noise in the signal, retain the core characteristics of the shaft system torsional vibration signal, and make the reconstructed signal clearer and more accurate. Moreover, in the process of constructing the shaft system torsional vibration modal function, the characteristics of the modal function are carefully screened. This screening process is based on preset conditions, ensuring that the selected modal function can accurately reflect the characteristics of the shaft system torsional vibration, enhancing the recognition of the features, and providing strong support for subsequent feature extraction and early warning judgment. In addition, by constructing an accurate shaft system torsional vibration modal function, the operating status of the shaft system can be more accurately reflected. When the shaft system is abnormal, such as the torsional vibration signal characteristics deviating from the normal range, the early warning system can issue a warning signal in a timely and accurate manner, thereby improving the reliability and real-time performance of the early warning system. Finally, the characteristics of the modal function are screened to reduce false alarms caused by signal noise or other interference factors. At the same time, since the selected modal function can accurately reflect the characteristics of the shaft system torsional vibration, it also reduces the missed alarm rate caused by inaccurate feature extraction.

[0051] Optionally, the modal reconstruction of the noise-reduced shaft torsional vibration signal to obtain a modal function includes: reconstructing the modal of the noise-reduced shaft torsional vibration signal based on morphological opening operation to obtain a morphological opening operation function; reconstructing the modal of the noise-reduced shaft torsional vibration signal based on morphological closing operation to obtain a morphological closing operation function; and obtaining the modal function based on the morphological opening operation function and the morphological closing operation function. In this embodiment, the peak noise and the concave noise in the signal can be effectively removed respectively through the morphological opening operation and the closing operation, thereby enhancing the effective features in the signal. Moreover, the morphological opening operation and the closing operation can retain the main shape characteristics of the signal while suppressing noise interference and improving the accuracy of signal analysis. In addition, the signal after morphological processing is more concise and clear in representation, which helps to more accurately describe the dynamic characteristics of the shaft torsional vibration. Finally, by morphologically reconstructing the noise-reduced shaft torsional vibration signal, a more stable and reliable modal function can be constructed, which helps to improve the resistance of the early warning system to noise and interference, and enhance the robustness and reliability of the early warning system.

[0052] Optionally, based on the following morphological opening operation formula, the noise reduction shaft system torsional vibration signal modal is reconstructed to obtain a morphological opening operation function: in, is the morphological opening function, c j * (n) is the approximate signal after peak removal, Detail signal after peak removal. In this embodiment, the morphological opening operation can remove the spike noise in the signal, making the reconstructed signal smoother and improving the signal-to-noise ratio of the signal. Moreover, while removing the noise, the morphological opening operation can better preserve the geometric shape and structural characteristics of the signal, avoiding the signal distortion problem that may be caused by traditional filtering methods. In addition, the morphological opening operation performs local processing of the signal based on the structural elements. Compared with the global processing algorithm, its computational efficiency is higher, and it can quickly complete the reconstruction and feature extraction of the signal, which is suitable for real-time early warning systems. Finally, the morphological opening operation is insensitive to slight changes in the signal, and can resist interference and noise in the signal to a certain extent, thereby improving the robustness and stability of the entire early warning system.

[0053] Optionally, based on the following morphological closed operation formula, the noise reduction shaft system torsional vibration signal modal is reconstructed to obtain a morphological closed operation function: is the morphological closing function, c j * (n) is the approximate signal after valley removal, d j #(n) is the detail signal after valley removal. In this embodiment, the concave noise is removed. The morphological closing operation effectively removes the concave noise in the signal by filling the concave part (i.e., local minimum or negative pulse) in the signal, thereby improving the signal quality and the accuracy of signal analysis. Moreover, the signal trend is retained. While removing the concave noise, the morphological closing operation can better retain the overall trend and main features of the signal, avoiding signal distortion caused by excessive smoothing. In addition, the reliability of the warning signal is enhanced. The modal function of the denoised shaft torsional vibration signal processed by the morphological closing operation is closer to the real shaft torsional vibration characteristics, thereby improving the accuracy and reliability of the warning signal. Finally, the morphological closing operation can flexibly perform noise suppression and signal reconstruction by adjusting the size and shape of the structural elements according to the different scales and characteristics of the noise in the signal, thereby enhancing the adaptability and flexibility of the method.

[0054] Optionally, the characteristics of the modal function are screened to construct a shaft system torsional vibration modal function based on the modal function that meets preset conditions, including: setting a frequency range based on the instantaneous frequency of the modal function; calculating the statistical characteristics of the instantaneous frequency; setting a screening standard based on the frequency range and the statistical characteristics; selecting the modal function that meets the conditions through the screening standard to construct a shaft system torsional vibration modal function; the statistical characteristics include the instantaneous frequency mean difference and the instantaneous frequency variance. In this embodiment, by setting the frequency range based on the instantaneous frequency of the modal function, it is possible to ensure that the analyzed signal frequency matches the actual frequency range of the shaft system torsional vibration, avoid interference from irrelevant frequency components, and improve the accuracy of signal processing. Moreover, calculating the statistical characteristics of the instantaneous frequency (such as the instantaneous frequency mean difference and the instantaneous frequency variance) can reflect the stability and fluctuation of the signal, thereby more accurately identifying abnormal conditions of the shaft system torsional vibration and enhancing the sensitivity of fault detection. In addition, the screening criteria set based on frequency range and statistical characteristics can eliminate modal functions that do not meet the requirements and retain only those functions that can truly reflect the torsional vibration characteristics of the shaft system, thereby improving the reliability and accuracy of the entire early warning system. Finally, by extracting and comparing the features of the screened modal functions, it is possible to more accurately determine whether the shaft system torsional vibration is in an abnormal state, thereby issuing a more accurate early warning signal and reducing false alarms and missed alarms.

[0055] The shaft system torsional vibration modal function is a modal function that can reflect the shaft system torsional vibration characteristics.

[0056] Optionally, the instantaneous frequency of the modal function is calculated based on the following formula, so as to set the frequency range based on the instantaneous frequency of the modal function: Among them, when the opening operation is used When the closing operation is used is the result of Hilbert transform, ∠ represents the phase angle, Represents the change in phase angle between adjacent time points, and Δn represents the time interval between phase time points. In this embodiment, the phase information is obtained by Hilbert transform, and the change in phase angle between adjacent time points is calculated to obtain an accurate instantaneous frequency. Moreover, the formula distinguishes between the two cases of morphological opening operation and closing operation, and gives the corresponding instantaneous frequency calculation methods respectively. This distinction enables the method to adapt to different types of noise and signal characteristics more flexibly, improving the adaptability and accuracy of the algorithm. In addition, time-frequency is an important parameter in signal analysis, which reflects the frequency characteristics of the signal at a certain moment. By setting the frequency range, the modal function that meets the specific frequency characteristics can be further screened out, thereby more accurately reflecting the actual situation of the torsional vibration of the shaft system. Finally, setting the screening criteria based on the statistical characteristics of the instantaneous frequency (such as mean and variance) can effectively eliminate the interference of outliers or noise and improve the robustness of the algorithm. At the same time, by adjusting the screening criteria, the strictness of the screening can be flexibly controlled to adapt to different application scenarios.

[0057] Optionally, the frequency range is set according to the instantaneous frequency of the modal function based on the following formula: min ≤f j (n)≤f max Among them, f min is the set minimum instantaneous frequency, f max is the maximum instantaneous frequency set, f j (n) is the instantaneous frequency. In the present embodiment, by setting the minimum instantaneous frequency and the maximum instantaneous frequency, the upper and lower limits of the instantaneous frequency are clarified, which provides a clear boundary division for subsequent analysis and processing of signals. Moreover, clarifying the frequency range helps to reduce misjudgment, avoids incorporating non-critical or unnecessary signal components into the analysis, and improves the accuracy of signal processing. In addition, by setting the frequency range, it is easier to filter out the signal segment of interest from a large amount of data based on this screening condition, reducing the processing difficulty and calculation complexity. Finally, by combining the set frequency range with the early warning mechanism, abnormal signals can be detected and identified within a specific frequency range, making the early warning mechanism more accurate and timely.

[0058] Optionally, the mean of the instantaneous frequency is calculated based on the following formula: in, is the mean instantaneous frequency, f j(n) is the instantaneous frequency, and N is the instantaneous frequency number. In this embodiment, by averaging all instantaneous frequencies, the formula can provide a more accurate and comprehensive frequency representation, rather than relying solely on a single instantaneous frequency value. This helps to more accurately describe the overall frequency characteristics of the shaft system torsional vibration signal over a period of time. Moreover, the mean calculation can reduce the sharp changes in instantaneous frequency caused by random fluctuations or noise in the signal, making the results more stable and reliable. This is very important for subsequent signal analysis and early warning decisions. In addition, the calculated instantaneous frequency mean can be used as the basis for subsequent signal processing and analysis, such as setting screening criteria, feature comparison, etc. It helps to further explore the potential characteristics of the signal and improve the accuracy and reliability of the early warning system. Finally, this formula is applicable to the analysis of shaft system torsional vibration signals at different scales and under different conditions. By calculating the instantaneous frequency mean of different signals, the frequency characteristics under different conditions can be obtained, thereby providing strong support for early warning systems in different application scenarios.

[0059] Optionally, the variance of the instantaneous frequency is calculated based on the following formula: where σ j is the variance of the instantaneous frequency, f j (n) is the instantaneous frequency, and N is the instantaneous frequency. In this embodiment, the instantaneous frequency variance is used as a quantitative indicator to evaluate the stability of the shaft system torsional vibration signal, which can accurately reflect the fluctuation of the signal in the time domain. A high variance value indicates that the signal fluctuates greatly, which may indicate an unstable state of the shaft system torsional vibration; a low variance value indicates that the signal is relatively stable. Moreover, by comparing the variance of the instantaneous frequency with the preset threshold, the early warning system can more accurately determine whether the shaft system torsional vibration is in an abnormal state, thereby improving the accuracy of the early warning. In addition, the calculation of the instantaneous frequency variance is combined with multi-scale morphological analysis to comprehensively evaluate the characteristics of the shaft system torsional vibration signal at different scales. Finally, based on the calculation of the instantaneous frequency variance, the early warning system can achieve real-time monitoring and rapid response. Once abnormal fluctuations in the shaft system torsional vibration signal are found, an early warning signal is immediately issued to provide timely guidance for equipment maintenance and troubleshooting.

[0060] Optionally, the screening criteria are set according to the mean value of the instantaneous frequency based on the following formula: Among them, τ is the screening criterion, f j (n) is the instantaneous frequency, is the mean of the instantaneous frequency. In this embodiment, by setting the screening criteria, it is clarified which modal functions meet the preset conditions, so that a shaft torsional vibration modal function that can accurately reflect the shaft torsional vibration characteristics can be constructed. This clarity helps to reduce subjective judgment and improve the reliability and consistency of the method. Moreover, the adjustment factor in the screening criteria (such as in the formula) allows adjustment according to specific application scenarios and requirements to adapt to different shaft torsional vibration signal characteristics. In addition, the screening criteria based on the mean of the instantaneous frequency can quickly screen out modal functions that meet the conditions, reduce the processing and analysis time of unnecessary data, and improve the efficiency of the overall method. Finally, the instantaneous frequency is an important parameter that reflects the dynamic characteristics of the signal. The screening criteria based on the mean of the instantaneous frequency can more accurately capture the key features in the shaft torsional vibration signal, thereby constructing a more accurate shaft torsional vibration modal function.

[0061] Optionally, the screening criterion is set according to the variance of the instantaneous frequency based on the following formula: σ j <τ, where τ is the screening criterion, σ j is the variance of the instantaneous frequency. In this embodiment, by calculating the variance of the instantaneous frequency and setting the screening criteria, the modal functions that meet the specific conditions can be screened out more accurately. This method can exclude those signals with excessive frequency fluctuations, which often represent unstable or abnormal torsional vibration states, thereby improving the accuracy of the early warning. Moreover, the method based on variance screening usually has a faster calculation speed, can complete the processing and analysis of a large amount of data in a shorter time, and improve the real-time nature of the early warning.

[0062] Optionally, the feature extraction of the modal function is performed to compare the features with the set modal threshold, and the early warning device issues an early warning signal based on the comparison result, including: extracting the features of the modal function to obtain feature information, the feature information includes: instantaneous frequency, modal average instantaneous frequency, modal energy, modal peak value, modal kurtosis; based on the feature information, feature comparison is performed with the set modal threshold, and the early warning device issues an early warning signal based on the comparison result, and the modal threshold includes: average instantaneous frequency threshold, energy threshold, peak value threshold, kurtosis threshold. In this embodiment, the state of the shaft system torsional vibration signal is comprehensively analyzed by extracting a variety of feature information (such as instantaneous frequency, modal average instantaneous frequency, modal energy, modal peak value, modal kurtosis). These features cover the frequency characteristics, energy distribution, peak characteristics and morphological characteristics of the signal, so that the health status of the shaft system can be comprehensively evaluated from multiple dimensions. Moreover, by setting precise modal thresholds (such as average instantaneous frequency threshold, energy threshold, peak value threshold, kurtosis threshold), it is possible to accurately determine whether the shaft system torsional vibration signal is abnormal. These thresholds are set based on empirical values ​​or system-specific conditions, which improves the accuracy of early warning signals.

[0063] Optionally, the modal instantaneous frequency is calculated based on the following formula: Among them, when the opening operation is used When the closing operation is used is the result of Hilbert transform, ∠ represents the phase angle, Represents the change in phase angle at adjacent time points, and Δn represents the time interval between phase time points. In this embodiment, by performing Hilbert transform on the de-noised shaft torsional vibration signal, the phase information of the signal can be accurately extracted, thereby calculating the instantaneous frequency. This method of calculating instantaneous frequency based on phase change is more accurate than the traditional method based on time domain or frequency domain analysis because it takes into account the dynamic change of the signal over time. Moreover, the formula is defined for the results of morphological opening and closing operations respectively, and can flexibly adapt to signals after different noise removal processes. The opening operation is suitable for removing peak noise, and the closing operation is suitable for removing concave noise. After processing them separately, the instantaneous frequency is calculated, which can more accurately reflect the true situation of shaft torsional vibration. In addition, this formula is not only applicable to the processing of shaft torsional vibration signals, but can also be extended to other similar fields where the instantaneous frequency needs to be calculated. As long as the signal has periodic or changing characteristics, the formula can be used for analysis. Finally, the calculation of the instantaneous frequency is based on the real-time phase change of the signal, which can realize real-time monitoring and early warning of the shaft torsional vibration state. This is particularly important for monitoring systems that require high real-time requirements, and potential problems can be discovered in time and measures can be taken to deal with them.

[0064] Optionally, the modal average instantaneous frequency is calculated based on the following formula: in, is the mean instantaneous frequency, f j (n) is the instantaneous frequency, and N is the instantaneous frequency. In this embodiment, by calculating the modal average instantaneous frequency, the characteristics of the shaft system torsional vibration signal can be quantitatively analyzed. Moreover, the modal average instantaneous frequency is an important indicator of the overall dynamic performance of the signal. By calculating this indicator, the overall stability and working condition of the shaft system torsional vibration signal can be evaluated to help identify potential faults or abnormalities. In addition, the calculated modal average instantaneous frequency can be compared with a preset threshold to determine whether the shaft system torsional vibration signal is within the normal range. Finally, during the debugging and maintenance of the equipment, the calculation results of the modal average instantaneous frequency can provide a basis for parameter adjustment.

[0065] Optionally, the modal energy is calculated based on the following formula: Among them, E j is the modal energy. When the open operation is used Or, when using the closing operation In this embodiment, by considering the signal after morphological opening and closing operations respectively to calculate the modal energy, the energy distribution of the signal under different processing states can be more accurately reflected. The morphological opening operation is mainly used to remove the peak noise in the signal, while the closing operation is used to remove the concave noise. Such processing makes the calculated modal energy closer to the real energy state of the signal.

[0066] Optionally, the modal peak is calculated based on the following formula: Among them, P j The modal peak value is when the open operation is used Or, when using the closing operation In this embodiment, by considering the signals after morphological opening and closing operations respectively to calculate the modal peak, the peak characteristics of the signal under different processing states can be captured more accurately. This differentiated processing helps to analyze the signal characteristics more carefully and improve the accuracy of the early warning system. Moreover, the formula is clearly expressed, the calculation process is relatively simple, and it is easy to implement in practical applications. This helps to reduce the complexity and cost of the early warning system and improve the reliability and stability of the system.

[0067] Optionally, the modal kurtosis is calculated based on the following formula: Among them, K j is the modal kurtosis, when the opening operation is used Or, when using the closing operation N is the number of morphological closed functions. In this embodiment, by providing formulas for both open and closed operations, the method can flexibly cope with different types of noise and signal characteristics. The open operation is mainly used to remove peak noise, while the closed operation is used to fill the concave noise. This dual processing method makes the method more comprehensive and flexible. Moreover, kurtosis is an important statistic that describes the sharpness of the signal waveform, and is very effective for detecting abnormal phenomena such as shocks and mutations in the signal. By calculating the kurtosis of the signal after morphological processing, these features in the shaft system torsional vibration signal can be more accurately reflected, thereby improving the accuracy and reliability of the early warning. In addition, morphological processing itself has the ability of denoising and feature enhancement. By preprocessing the signal through morphological open and closed operations, the influence of noise on the kurtosis calculation can be significantly reduced, making the early warning system more robust to noise and interference. Finally, morphological processing is combined with multi-scale analysis methods such as wavelet decomposition, so that the method can perform detailed analysis of the signal at different scales. By calculating the modal kurtosis at different scales, the characteristics of the shaft system torsional vibration signal can be more comprehensively understood, providing richer information support for early warning.

[0068] Optionally, the average instantaneous frequency threshold is set based on the following formula: Among them, fthreshold is the instantaneous frequency threshold, K is the adjustment factor, and Δt is the update period. In this embodiment, by introducing the adjustment factor α and the update period T, this embodiment allows the average instantaneous frequency threshold to be dynamically adjusted according to the actual operating conditions. Moreover, since the average instantaneous frequency is one of the important characteristics of the shaft system torsional vibration signal, accurately setting its threshold is crucial for timely detection of potential problems. This embodiment calculates and adjusts the threshold so that the early warning system can more accurately capture abnormal changes in the shaft system torsional vibration signal. Optionally, the energy threshold is set based on the following formula: E threshold (t+Δt)=E threshold (t)+k*(E j (t)-E threshold (t)), where E threshold is the energy threshold, K is the adjustment factor, and Δt is the update cycle. In this embodiment, by introducing the adjustment factor (α) and the update cycle (T), the formula makes the energy threshold no longer statically fixed, but can be dynamically adjusted according to the actual situation. This dynamic adjustment mechanism makes the early warning system more flexible, able to adapt to energy changes under different working conditions and environments, and improve the accuracy and reliability of the early warning. Moreover, the energy characteristics of different shaft torsional vibration systems or different working conditions may be different. Through the adjustment factor, the energy threshold can be fine-tuned according to the actual monitoring data and system characteristics to ensure that the threshold setting is closer to the actual situation and reduce false alarms and missed reports. In addition, the setting of the update cycle (T) allows the system to regularly update the energy threshold based on the latest monitoring data, which helps to maintain the timeliness and accuracy of the early warning system. As the system runs for a longer time, its energy characteristics may change. Regularly updating the threshold can ensure that the system is always in the best early warning state. Finally, a reasonable energy threshold setting can improve the sensitivity of the early warning system to abnormal energy. Once the energy is detected to exceed the normal range, a warning signal can be issued in time to avoid potential faults from causing greater damage to the system.

[0069] Optionally, the peak threshold is set based on the following formula: threshold (t+Δt)=P threshld (t)+k*(P j (t)-P threshold (t)) where P threshold is the peak threshold, K is the adjustment factor, and Δt is the update period. In this embodiment, the adjustment factor in the formula allows the user to adjust the peak threshold according to the specific situation to adapt to different shaft torsional vibration signal characteristics and early warning requirements. This flexibility improves the application scope and accuracy of the method. Moreover, by introducing the update period, the peak threshold can be dynamically adjusted according to the change of the signal instead of being fixed. This dynamic adaptability helps to maintain the effectiveness and accuracy of the early warning system when the signal characteristics change.

[0070] Optionally, the kurtosis threshold is set based on the following formula: K threshold (t+Δt)=K threshold (t)+k*(K j (t)-K threshold (t)), where K threshold is the kurtosis threshold, K is the adjustment factor, and Δt is the update period. In this embodiment, the adjustment factor (α) and the update period (T) are introduced into the formula, so that the kurtosis threshold (K_threshold) can be dynamically adjusted according to the actual operating conditions and needs. This flexibility helps to adapt to the changes in the shaft system torsional vibration signal under different working conditions and improve the adaptability and accuracy of the early warning system. Moreover, kurtosis is a statistic that reflects the sharpness of the signal distribution morphology, and is very sensitive to detecting abnormal peaks or depressions in the shaft system torsional vibration signal. By setting a reasonable kurtosis threshold, abnormal changes in the shaft system torsional vibration signal can be more effectively identified, thereby improving the accuracy and reliability of the early warning. In addition, dynamically adjusting the kurtosis threshold can reduce false alarms or missed alarms caused by improper setting of the fixed threshold. During the operation of the system, adjusting the threshold according to the actual signal characteristics helps to enhance the robustness and stability of the system and improve the overall early warning performance. Finally, the dynamic adjustment of the kurtosis threshold also helps system maintenance personnel understand the changing trends and characteristics of the shaft system torsional vibration signal, providing an important reference for system maintenance and optimization.

[0071] Optionally, the feature information is compared with the set modal threshold, and the early warning device issues an early warning signal based on the comparison result, including: if at least any one of the following conditions is met, the early warning device issues an early warning signal: the modal average instantaneous frequency is greater than the average instantaneous frequency threshold; the modal energy is equal to the energy threshold; the modal peak is equal to the peak threshold; the modal kurtosis is equal to the kurtosis threshold. In this embodiment, the specific conditions for triggering the early warning signal are clearly pointed out, that is, when the four characteristic values ​​of the modal average instantaneous frequency, modal energy, modal peak and modal kurtosis exceed or are equal to their corresponding thresholds, the early warning device will issue an early warning signal. This clarity helps the user or system to clearly understand when the early warning will be triggered, which is convenient for fault prevention and the formulation of countermeasures. Moreover, by setting different thresholds, this embodiment provides high flexibility. Users can adjust these thresholds to meet different monitoring needs according to specific application scenarios and the characteristics of the shaft system torsional vibration signal.

[0072] This embodiment also provides a shaft system torsional vibration signal warning method based on a multi-scale morphological algorithm, including: a memory and a processor, the memory storing a computer executable program, and the processor being used to execute the computer executable program to implement any of the methods described in the embodiments.

[0073] This embodiment further provides a shaft system torsional vibration signal warning method based on a multi-scale morphological algorithm. A computer executable program is stored on the computer storage medium, and when the computer executable program is executed, any method described in the embodiment is implemented.

[0074] It should be noted that the same and similar parts between the various embodiments in this specification can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments. The device and system embodiments described above are merely schematic, wherein the modules described as separate components may or may not be physically separated, and the components indicated as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. A person of ordinary skill in the art can understand and implement it without paying creative labor.

[0075] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A shaft torsional vibration signal early warning method based on a multi-scale morphological algorithm, characterized in that: include: Preprocessing the original shaft torsional vibration signal collected by the sensor to obtain the shaft torsional vibration time series signal; Performing feature decomposition on the shaft system torsional vibration time series signal to obtain an abnormal shaft system torsional vibration signal; Performing morphological processing on the abnormal shafting torsional vibration signal to obtain a noise-reduced shafting torsional vibration signal; Processing the noise-reduced shafting torsional vibration signal modally to construct a shafting torsional vibration modal function; The modal function is subjected to feature extraction to obtain the shaft system torsional vibration feature, and the feature is compared with the set modal threshold to generate an early warning signal.

2. The shafting torsional vibration signal early warning method based on multi-scale morphological algorithm according to claim 1 is characterized in that: The characteristic decomposition of the shaft system torsional vibration time sequence signal to obtain an abnormal shaft system torsional vibration signal includes: Performing wavelet decomposition on the shaft system torsional vibration time series signal to obtain approximate signals and detail signals at different scales; Statistical characteristic analysis is performed on the approximate signal and the detail signal to obtain an abnormal shafting torsional vibration signal.

3. The shafting torsional vibration signal early warning method based on multi-scale morphological algorithm according to claim 2 is characterized in that: Based on the following formula, the shaft system torsional vibration time series signal is decomposed by wavelet to obtain approximate signals and detail signals at different scales: Among them, c j,k is the wavelet coefficient of the approximate signal, d j,k is the wavelet coefficient of the detail signal, φ j,k (n) is the scale function, ψ j,k (n) is the wavelet function, j is the scale parameter, k is the translation parameter, J is the summation termination value, and n is the discrete time index.

4. The shafting torsional vibration signal early warning method based on multi-scale morphological algorithm according to claim 2 is characterized in that: The scaling function is calculated based on the following formula: f j,k (n)=2 -j / 2 φ(2 -j n.k.); Among them, φ j,k (n) is the scaling function, n is the discrete time index, k is the translation parameter, and j is the scaling parameter.

5. The shafting torsional vibration signal early warning method based on multi-scale morphological algorithm according to claim 2 is characterized in that: The wavelet function is calculated based on the following formula: Among them, ψ j,k (n) is the wavelet function, n is the discrete time index, k is the translation parameter, and j is the scale parameter. ψ j,k (n)=2 -j / 2 ψ(2 -j nk).

6. The shafting torsional vibration signal early warning method based on multi-scale morphological algorithm according to claim 2 is characterized in that: The performing morphological processing on the abnormal shafting torsional vibration signal to obtain a noise-reduced shafting torsional vibration signal includes: Performing a morphological opening operation on the peak noise in the approximate signal and the detail signal to remove the peak noise and obtain a de-peaked signal; Performing a morphological closing operation on the concave noise in the approximate signal and the detail signal to remove the concave noise and obtain a de-valleyed signal; A noise-reduced shaft system torsional vibration signal is obtained based on the de-peaked signal and the de-valleyed signal.

7. The shafting torsional vibration signal early warning method based on multi-scale morphological algorithm according to claim 1 is characterized in that: The noise reduction shafting torsional vibration signal is modally processed to construct a shafting torsional vibration modal function, including: Reconstructing the noise-reduced shaft system torsional vibration signal mode to obtain a modal function; The characteristics of the modal functions are screened to construct a shaft system torsional vibration modal function based on the modal functions that meet preset conditions.

8. The shafting torsional vibration signal early warning method based on multi-scale morphological algorithm according to claim 7 is characterized in that: The modal reconstruction of the noise reduction shafting torsional vibration signal to obtain a modal function includes: Reconstructing the noise reduction shaft system torsional vibration signal mode based on morphological opening operation to obtain a morphological opening operation function; Reconstructing the noise reduction shaft system torsional vibration signal mode based on morphological closing operation to obtain a morphological closing operation function; The modal function is obtained based on the morphological opening operation function and the morphological closing operation function.

9. The shafting torsional vibration signal early warning method based on multi-scale morphological algorithm according to claim 7 is characterized in that: The screening of the characteristics of the modal function to construct a shaft system torsional vibration modal function based on the modal function satisfying a preset condition comprises: setting a frequency range based on an instantaneous frequency of the modal function; calculating statistical characteristics of the instantaneous frequency; setting a screening criterion based on the frequency range and the statistical characteristics; The modal functions that meet the conditions are selected through the screening criteria to construct the shaft system torsional vibration modal functions.

10. The shafting torsional vibration signal early warning method based on multi-scale morphological algorithm according to claim 1 is characterized in that: Extracting features of the modal function to obtain shaft system torsional vibration features, and performing feature comparison with a set modal threshold to generate an early warning signal, including: Perform feature extraction on the modal function to obtain feature information, wherein the feature information includes: Instantaneous frequency, modal average instantaneous frequency, modal energy, modal peak, modal kurtosis; Based on the feature information, feature comparison is performed with the set modal threshold, where the modal threshold includes: average instantaneous frequency threshold, energy threshold, peak threshold, and kurtosis threshold; If at least one of the following conditions is met, an early warning signal is generated: The modal average instantaneous frequency is greater than the average instantaneous frequency threshold; The modal energy is equal to the energy threshold; The modal peak value is equal to the peak threshold value; The modal kurtosis is equal to the kurtosis threshold.