A method and system for identifying speed fluctuations
By sorting and percentile windowing the vibration signal spectrum of rotating machinery, a statistical baseline value is constructed. Combined with confidence assessment and iterative search, the problem of continuous tracking and online monitoring of frequency fluctuations in rotating machinery is solved, and efficient and accurate frequency identification is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- TANGZHI SCI & TECH HUNAN DEV CO LTD
- Filing Date
- 2026-05-15
- Publication Date
- 2026-06-12
AI Technical Summary
Existing technologies struggle to continuously track frequency fluctuations in rotating machinery and have high computational complexity, failing to meet the demands for online real-time monitoring.
By acquiring the vibration signal spectrum of rotating machinery, sorting the amplitudes, determining the percentile window range, constructing a statistical baseline value, using this baseline value to identify frequency fluctuations, and employing a weighted average based on location index and confidence assessment, combined with a bilateral iterative search, to determine the frequency boundary.
It achieves continuous tracking of frequency fluctuations, has strong anti-interference capabilities, low computational complexity, is suitable for online real-time monitoring, and improves the accuracy and robustness of identification.
Smart Images
Figure CN122192502A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of rotating machinery condition monitoring technology, and in particular relates to a method and system for identifying speed fluctuations. Background Technology
[0002] During operation, especially under acceleration, deceleration, or variable load conditions, the rotational frequency of rotating machinery fluctuates over time. Accurately obtaining the real-time trend of frequency changes is crucial for equipment condition monitoring, fault diagnosis, and speed control. In practical engineering, many rotating machines lack speed sensors, or the acquired speed signals are of poor accuracy, providing only rough reference values. Therefore, it is often necessary to directly identify the rotational frequency and its fluctuations from vibration signals.
[0003] One traditional method is spectral peak search: in the spectrum of the vibration signal, the maximum spectral line is searched for nearby based on the reference frequency, and the corresponding frequency is taken as the frequency. This method is simple to implement, but it can only obtain a single frequency value and cannot describe the continuous fluctuation process of the frequency over time. Another method is to use high-resolution time-frequency analysis technology, such as continuous wavelet transform. This method can extract the trajectory of frequency change over time, but the computational load is large and it is difficult to meet the needs of online real-time monitoring. Summary of the Invention
[0004] In view of this, this application provides a method and system for identifying rotational speed fluctuations, which can continuously track the frequency fluctuation process, has strong anti-interference ability, does not rely on subjective thresholds, and has low computational complexity, making it suitable for online real-time monitoring.
[0005] This application provides a method for identifying speed fluctuations, the method comprising: Obtain the spectrum of vibration signals from rotating machinery; The amplitude values in the spectrum are sorted to obtain an ordered spectrum; In the ordered spectrum, a percentile window is determined by a first high percentile and a second high percentile below the first high percentile, and the amplitude range corresponding to the percentile window is located. Based on each amplitude within the stated amplitude range, a statistical baseline value is obtained; The frequency of the vibration signal is identified using the statistical baseline value.
[0006] Optionally, the first high percentile is 85% to 95%, and the second high percentile is 10% to 20% lower than the first high percentile.
[0007] Optionally, obtaining the statistical baseline value based on each amplitude within the amplitude range includes: A weighted average based on the location index is performed on each amplitude within the amplitude range to obtain a weighted average result, and the weighted average result is used as the statistical baseline value.
[0008] Optionally, the weights used in the weighted average are calculated by using a Gaussian function to decay the distance between the location index and the center position of the amplitude range, or by using a linear decay to decay the distance between the location index and the center position of the amplitude range.
[0009] Optionally, identifying the frequency of the vibration signal using the statistical baseline value includes: Obtain the reference frequency; Spectral peaks are searched within the neighborhood of the reference frequency, and the confidence level of each spectral peak is obtained using the statistical baseline value, the amplitude of each spectral peak, the standard deviation of the spectral amplitude within the neighborhood of the reference frequency, the frequency of the spectral peak, the reference frequency, and the frequency offset penalty coefficient. The frequency corresponding to the spectral peak of the maximum value among the various confidence levels is determined as the initially identified frequency.
[0010] Optionally, the confidence level is jointly determined by the amplitude difference factor and the frequency offset factor; The amplitude difference factor is used to assess the significance of the amplitude of the spectral line peak based on the deviation between the amplitude of the spectral line peak and the statistical baseline value, as well as the standard deviation of the spectral line amplitude in the neighborhood of the reference frequency. The frequency offset factor is used to evaluate the frequency matching degree of the spectral peak based on the deviation between the frequency of the spectral peak and the reference frequency, as well as the frequency offset penalty coefficient.
[0011] Optionally, it also includes: The reliability index of the initially identified frequency is obtained based on the amplitude of the spectral line corresponding to the reference frequency, the statistical baseline value, the maximum amplitude of the spectral line in the neighborhood of the reference frequency, the variance of the spectral line amplitude in the local neighborhood centered on the initially identified frequency, and the variance of the spectral line amplitude in the neighborhood of the reference frequency. The reliability index is used to evaluate the reliability of the initially identified frequency.
[0012] Optionally, it also includes: Using the initially identified frequency as the center frequency, a bilateral iterative search is performed in both the low-frequency and high-frequency directions to determine the low-frequency and high-frequency boundaries of the frequency fluctuation.
[0013] Optionally, the bilateral iterative search includes: In each iteration, determine the local maximum point in the current local neighborhood of the current frequency point; Based on the amplitude of each frequency point in the current local neighborhood, the mean amplitude in the current local neighborhood, and the standard deviation of the amplitude in the current local neighborhood, the statistical significance index of each frequency point in the current local neighborhood is obtained. When the maximum value of the statistical significance index of each local maximum point in the current local neighborhood is less than a preset threshold, and the statistical significance index of each of the preset number of consecutive frequency points in the search direction is less than a preset proportion of the preset threshold, the search is stopped, and the current frequency is determined as the boundary in the current search direction.
[0014] This application also provides a speed fluctuation identification system, the system comprising: The acquisition module is used to acquire the spectrum of vibration signals from rotating machinery. The sorting module is used to sort the amplitude values in the spectrum to obtain an ordered spectrum; The positioning module is used to determine, in the ordered spectrum, a percentile window defined by a first high percentile and a second high percentile lower than the first high percentile, and to locate the amplitude range corresponding to the percentile window; The processing module is used to obtain a statistical baseline value based on each amplitude within the amplitude range; The identification module is used to identify the frequency of the vibration signal using the statistical baseline value.
[0015] Compared with the prior art, the speed fluctuation identification method and system provided in this application have the following advantages: (1) It can realize continuous tracking of frequency fluctuation process: This application constructs a dynamic baseline based on the statistical distribution of spectrum and uses the baseline to assist in frequency identification. It can effectively extract the frequency change trend over time, overcoming the shortcomings of traditional spectrum peak search methods that can only obtain a single frequency value and cannot describe the speed fluctuation process.
[0016] (2) Strong anti-interference ability, no need to rely on subjective threshold: This application calculates the statistical baseline by sorting the spectral amplitude and selecting the amplitude within the high percentile window. Based on the distribution characteristics of the data itself, the baseline can adaptively suppress the interference of background noise and abnormal spectral lines. Compared with the traditional method that relies on manually setting thresholds, it has higher stability and robustness.
[0017] (3) Low computational complexity, suitable for online real-time monitoring: This application mainly involves basic operations such as spectrum sorting and percentile statistics, without the need for complex time-frequency transformation (such as continuous wavelet transform), and the amount of computation is small, which can meet the real-time requirements of online condition monitoring of rotating machinery. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 This is a flowchart illustrating a speed fluctuation identification method disclosed in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a speed fluctuation identification system disclosed in an embodiment of this application. Detailed Implementation
[0020] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0021] It should be noted that when a component is referred to as being "fixed to" or "set on" another component, it can be directly on or indirectly set on the other component; when a component is referred to as being "connected to" another component, it can be directly connected to or indirectly connected to the other component.
[0022] It should be noted that the structures, proportions, sizes, etc., shown in the accompanying drawings of this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed in the specification, and are not intended to limit the conditions under which this application can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size should still fall within the scope of the technical content disclosed in this application, provided that they do not affect the effects and purposes that this application can produce.
[0023] like Figure 1 As shown in the figure, this application provides a method for identifying speed fluctuations, the method including: S11. Obtain the spectrum of the vibration signal of the rotating machinery; In this embodiment, vibration signals can be collected by an accelerometer installed on rotating machinery, and the time-domain vibration signals can be converted into frequency-domain spectra using fast Fourier transform. The spectrum reflects the energy distribution of different frequency components, providing basic data for subsequent frequency conversion identification.
[0024] S12. Sort the amplitude values in the spectrum to obtain an ordered spectrum; In this embodiment, the sorting operation can be represented as: ; Where sort is the sorting function; The amplitude in the spectrum; This is an ordered spectrum arranged from smallest to largest amplitude.
[0025] In this embodiment, the amplitudes of all frequency points in the spectrum can be rearranged in ascending order to form an ordered spectrum. This sorting operation eliminates the influence of the original frequency order in the spectrum, allowing subsequent statistical analysis to be based on the overall distribution of amplitudes rather than frequency positions, thereby enhancing robustness to outliers.
[0026] S13. In the ordered spectrum, determine the percentile window defined by the first highest percentile and the second highest percentile below the first highest percentile, and locate the amplitude range corresponding to the percentile window. In this embodiment, the first high percentile is used to select the region with a large amplitude in the ordered spectrum, and the second high percentile is used to define the lower boundary of the window. By using these two percentiles, a continuous amplitude range at the high end of the ordered spectrum can be dynamically extracted. This range represents the amplitude level of the main energy components in the signal, which can effectively avoid the influence of background noise and extreme peaks.
[0027] S14. Obtain the statistical baseline value based on each amplitude within the amplitude range; In this embodiment, statistical calculations are performed on all amplitudes within the amplitude range corresponding to the percentile window to obtain a statistical baseline value characterizing the main energy level of the spectrum. Specifically, this can be achieved by performing a weighted average based on the position index on each amplitude within the amplitude range, and using the weighted average result as the statistical baseline value. Alternatively, the median of the amplitudes within the amplitude range can be taken as the statistical baseline value. Another option is to sort all amplitudes within the amplitude range by size, remove the largest 10% and the smallest 10% of amplitudes, calculate the mean of the remaining amplitudes, and use this mean as the statistical baseline value.
[0028] This baseline value reflects the normal amplitude reference of the vibration signal under conditions of no significant interference, providing a stable comparison benchmark for accurate identification of subsequent frequency shifts.
[0029] S15. Use statistical baseline values to identify the frequency of vibration signals.
[0030] In this embodiment, a statistical baseline is used as a reference, combined with the reference rotational frequency of the rotating machinery, to identify the rotational frequency components and their fluctuation ranges in the spectrum. Specifically, the statistical baseline is used to assess the significance of each frequency component to distinguish the true rotational frequency from background noise, thereby achieving accurate positioning of the rotational frequency and extraction of its fluctuation boundaries. Because the statistical baseline can resist local noise and interference spectral peaks, the identification results are more accurate and reliable, and it does not require complex time-frequency transformations, has low computational load, and is suitable for online monitoring.
[0031] Compared with the prior art, the speed fluctuation identification method and system provided in this application have the following advantages: (1) It can realize continuous tracking of frequency fluctuation process: This application constructs a dynamic baseline based on the statistical distribution of spectrum and uses the baseline to assist in frequency identification. It can effectively extract the frequency change trend over time, overcoming the shortcomings of traditional spectrum peak search methods that can only obtain a single frequency value and cannot describe the speed fluctuation process.
[0032] (2) Strong anti-interference ability, no need to rely on subjective threshold: This application calculates the statistical baseline by sorting the spectral amplitude and selecting the amplitude within the high percentile window. Based on the distribution characteristics of the data itself, the baseline can adaptively suppress the interference of background noise and abnormal spectral lines. Compared with the traditional method that relies on manually setting thresholds, it has higher stability and robustness.
[0033] (3) Low computational complexity, suitable for online real-time monitoring: This application mainly involves basic operations such as spectrum sorting and percentile statistics, without the need for complex time-frequency transformation (such as continuous wavelet transform), and the amount of computation is small, which can meet the real-time requirements of online condition monitoring of rotating machinery.
[0034] In one implementation method, in this application embodiment, the first highest percentile is 85% to 95%, and the second highest percentile is 10% to 20% lower than the first highest percentile.
[0035] In this embodiment, the first high percentile is set to 85% to 95%, and the second high percentile is set to 10% to 20% lower than the first high percentile. This ensures that the selected percentile window is located in the region with higher amplitude in the ordered spectrum, thereby effectively capturing the main energy components in the vibration signal. At the same time, it avoids the influence of extreme outliers due to an excessively narrow window, or the introduction of too much background noise due to an excessively wide window.
[0036] As one implementation method, in this embodiment of the application, a statistical baseline value is obtained based on each amplitude within the amplitude range, including: A weighted average is calculated based on the location index for each amplitude within the amplitude range, and the weighted average result is used as the statistical baseline value.
[0037] In this embodiment, the statistical baseline value is calculated using a weighted average based on the location index, which can be obtained by the following formula: ; in, The baseline value is used for statistical analysis; k1 and k2 correspond to the position indices of the second and first highest percentiles in the ordered spectrum, respectively. The weight is for position k; It represents the k-th amplitude in the ordered spectrum.
[0038] In this embodiment, by performing a weighted average based on position index on each amplitude within the percentile window, the degree of contribution can be distinguished according to the relative position index of each amplitude within the window (e.g., amplitudes closer to the center of the window are given higher weights), making the statistical baseline value more reasonably reflect the energy concentration area of the main spectrum. Compared with simple averaging, weighted averaging can further suppress the influence of window edges or outliers, improving the stability and representativeness of the statistical baseline.
[0039] As one implementation method, in this embodiment of the application, the weighted average is calculated by a Gaussian function decay based on the distance between the position index and the center position of the amplitude range, or by a linear decay based on the distance between the position index and the center position of the amplitude range.
[0040] The weights are calculated by attenuating the distance between the location index and the center of the amplitude range using a Gaussian function, and can then be obtained using the following formula: ; in, The weight is the position k; k is the position index. The center position of the amplitude range; This is a decay rate control parameter, and its value range is typically [range missing]. .
[0041] The weights are calculated using a linear decay based on the distance between the location index and the center of the amplitude range, and can be obtained using the following formula: ; in, The weight is the position k; k is the position index. L represents the center position of the amplitude range; L is the width of the half-window, which is the distance from the center position to the window boundary.
[0042] In this embodiment, the weights calculated using a Gaussian function decay based on the distance between the position index and the center of the amplitude range are maximized at the window center and decay symmetrically to both sides, thus giving higher weights to amplitudes closer to the window center. Alternatively, the weights calculated using a linear decay based on the distance between the position index and the center of the amplitude range decrease linearly with increasing distance, resulting in less computation. Both weighting methods effectively reduce the impact of window boundary amplitudes on the statistical baseline value, improving the stability of the statistical baseline.
[0043] As one implementation method, in this embodiment of the application, step S15 includes: S151. Obtain the reference frequency; In this embodiment, the reference rotational frequency can be derived from the set rotational speed of the rotating machinery, the value collected by an external speed sensor, or the estimated rotational speed from historical operating data. This reference rotational frequency is used to limit the initial range of subsequent rotational frequency searches, thereby improving recognition efficiency.
[0044] S152. Search for spectral peaks in the neighborhood of the reference frequency, and use the statistical baseline value, the amplitude of each spectral peak, the standard deviation of the spectral amplitude in the neighborhood of the reference frequency, the frequency of the spectral peak, the reference frequency, and the frequency offset penalty coefficient to obtain the confidence level of each spectral peak. In this embodiment, confidence is calculated only for the peak values of the searched spectral lines within the neighborhood of the reference frequency. For each peak value, a comprehensive confidence score can be calculated based on the deviation of its amplitude from the statistical baseline, the standard deviation of the amplitude of the spectral lines within the neighborhood of the reference frequency, the deviation of its frequency from the reference frequency, and the frequency offset penalty coefficient. The smaller the deviation of the amplitude of the peak value from the statistical baseline, and the smaller the deviation of the frequency of the peak value from the reference frequency, the higher the confidence score of the peak value.
[0045] S153. The frequency corresponding to the peak value of the spectral line of the maximum value in each confidence level is determined as the initial identification frequency.
[0046] In this embodiment, the initially identified frequency switching frequency is determined by the following formula: ; in, Frequency switching for preliminary identification; The frequency of the spectral peak; Ω represents the amplitude of the spectral peak; Ω is the neighborhood of the reference frequency. ( Typically, the value is taken as 0.1-0.2). This is the confidence function.
[0047] In this embodiment, the frequency corresponding to the peak of the spectral line with the highest confidence level is selected as the initial identification frequency in the neighborhood of the reference frequency. This initial identification frequency utilizes both the robust estimation of the signal background by the statistical baseline value and the prior information of the reference frequency, which has a stronger anti-interference capability compared to simply searching for spectral peaks.
[0048] In one implementation, in this embodiment of the application, the confidence level is jointly determined by the amplitude difference factor and the frequency offset factor; wherein, the amplitude difference factor is used to evaluate the significance of the amplitude of the spectral peak based on the deviation between the amplitude of the spectral peak and the statistical baseline value, and the standard deviation of the spectral amplitude in the neighborhood of the reference frequency; the frequency offset factor is used to evaluate the frequency matching degree of the spectral peak based on the deviation between the frequency of the spectral peak and the reference frequency, and the frequency offset penalty coefficient.
[0049] In this embodiment, the amplitude difference factor assesses the significance of the spectral peak relative to the background energy by calculating the difference between the amplitude of the spectral peak and the statistical baseline value, and combining this with the standard deviation of the spectral amplitude in the neighborhood of the reference frequency. The larger the standard deviation, the more severe the amplitude fluctuation in the neighborhood, and the higher the tolerance for amplitude deviation. The frequency shift factor assesses the degree of matching between the frequency of the spectral peak and the expected frequency by calculating the deviation between the frequency of the spectral peak and the reference frequency, and combining this with the frequency shift penalty coefficient. The larger the shift penalty coefficient, the heavier the penalty for frequency deviation. The confidence level of the spectral peak is obtained by multiplying (or weighting) the amplitude difference factor and the frequency shift factor, so that spectral peaks with prominent amplitudes and frequencies close to the reference frequency receive higher confidence scores.
[0050] Specifically, the confidence function can take the following form: ; Where exp represents an exponential function with the natural constant e as its base; σ is the amplitude of the spectral peak; M is the statistical baseline value; σ is the standard deviation of the spectral amplitude in the neighborhood of the reference frequency; γ is the frequency offset penalty coefficient, which is used to control the penalty intensity when the frequency deviates from the reference frequency. Its value range can be 1-5, for example, the default value is 2. The frequency of the spectral peak; For reference frequency conversion.
[0051] Specifically, the standard deviation of the spectral line amplitudes in the neighborhood of the reference frequency can be obtained by the following formula: ; in, The standard deviation of the spectral line amplitude in the neighborhood of the reference frequency; The number of frequency points in the neighborhood of the reference frequency; Ω represents the frequency of the spectral line; Ω is the neighborhood of the reference frequency. The amplitude of the spectral line; The mean of the spectral line amplitudes in the neighborhood of the reference frequency.
[0052] Specifically, the mean value of the spectral line amplitude in the neighborhood of the reference frequency can be obtained by the following formula: ; in, The mean of the spectral line amplitudes in the neighborhood of the reference frequency; The number of frequency points in the neighborhood of the reference frequency; Ω represents the frequency of the spectral line; Ω is the neighborhood of the reference frequency. This represents the amplitude of the spectral line.
[0053] As one implementation method, this application embodiment further includes: S21. Based on the amplitude of the spectral line corresponding to the reference frequency, the statistical baseline value, the maximum amplitude of the spectral line in the neighborhood of the reference frequency, the variance of the spectral line amplitude in the local neighborhood centered on the initially identified frequency, and the variance of the spectral line amplitude in the neighborhood of the reference frequency, the reliability index of the initially identified frequency is obtained. The reliability index is used to evaluate the reliability of the initially identified frequency.
[0054] In this embodiment, the reliability index can be calculated using the following formula: ; in, As a reliability indicator, The closer the value is to 1, the more reliable the initial frequency identification is; The amplitude of the spectral line corresponding to the reference frequency is given; M is the statistical baseline value. Ω represents the frequency of the spectral line; Ω is the neighborhood of the reference frequency. The maximum amplitude of the spectral line in the neighborhood of the reference frequency; The variance of spectral line amplitudes within a local neighborhood centered on the initially identified frequency reversion point; The variance of spectral line amplitudes in the neighborhood of the reference frequency.
[0055] Specifically, the variance of the spectral line amplitude within a local neighborhood centered on the initially identified frequency transition can be obtained by the following formula: ; in, The variance of spectral line amplitudes within a local neighborhood centered on the initially identified frequency reversion point; The local neighborhood centered on the initially identified frequency switching frequency; The frequency of the spectral line; The amplitude of the spectral line; It represents the mean of the spectral line amplitudes within a local neighborhood centered on the initially identified frequency.
[0056] Specifically, the variance of the spectral line amplitude in the neighborhood of the reference frequency can be obtained by the following formula: ; in, The variance of spectral line amplitude in the neighborhood of the reference frequency; The number of frequency points in the neighborhood of the reference frequency; Ω represents the frequency of the spectral line; Ω is the neighborhood of the reference frequency. The amplitude of the spectral line; The mean of the spectral line amplitudes in the neighborhood of the reference frequency.
[0057] In this embodiment, the reliability index reflects the credibility of the initially identified frequency transition. Specifically, the energy significance and local stability of the initially identified frequency transition are comprehensively evaluated based on the difference between the amplitude of the spectral line corresponding to the reference frequency transition and the statistical baseline value, the difference between the amplitude of the maximum spectral line in the neighborhood and the baseline value, the variance of the spectral line amplitude in the local neighborhood of the initially identified frequency transition, and the variance of the global spectral line amplitude in the neighborhood of the reference frequency transition.
[0058] A higher reliability index value indicates that the initially identified frequency is energy-prominent and has stable local spectral fluctuations, with better consistency with the reference frequency's position, thus making the identification result more reliable. Conversely, a low reliability index suggests that the identification result may be affected by interference, requiring further verification or adjustment in subsequent steps. By introducing reliability assessment, misidentification caused by local spectral anomalies can be effectively avoided, improving the overall robustness of the method.
[0059] As one implementation method, this application embodiment further includes: S31. Using the initially identified frequency as the center frequency, perform bilateral iterative searches in both the low-frequency and high-frequency directions to determine the low-frequency and high-frequency boundaries of the frequency fluctuation.
[0060] In this embodiment, since the actual frequency shifts continuously during acceleration or deceleration, obtaining only a preliminary identified frequency point is insufficient to describe its fluctuation process. Therefore, using this preliminary frequency as the center, the search proceeds gradually towards both the low-frequency and high-frequency sides until the boundary position where the frequency component disappears or becomes insignificant is found. The interval defined by these two boundaries is the fluctuation range of the frequency shift over time. The bilateral iterative search does not require global analysis of the entire spectrum, resulting in high computational efficiency.
[0061] In this embodiment, the specific parameter settings for the bilateral iterative search are as follows: from the initially identified frequency conversion... Start with the preset step size (Typically 1-2Hz) It moves successively towards lower or higher frequencies, and in each iteration, it uses the current frequency point... Centered on a predetermined neighborhood radius (Typically 5-10Hz) Define the local neighborhood The step size and neighborhood radius can be adjusted according to the actual signal characteristics and accuracy requirements. Meanwhile, to prevent infinite searching due to spectral anomalies, an early termination protection mechanism is implemented, requiring low-frequency searches to not exceed [a certain threshold]. High-frequency boundary search does not exceed .
[0062] As one implementation method, in this embodiment of the application, the bilateral iterative search includes: S311. In each iteration, determine the local maximum point in the current local neighborhood of the current frequency point; In this embodiment, in each iteration, a local neighborhood (e.g., a frequency band of several hertz to the left and right) is defined centered on the currently examined frequency point. Within this neighborhood, the frequency point where the amplitude reaches a local maximum is searched. Local maxima are candidate locations for candidate frequency transition components. By focusing on these extreme points rather than all frequency points, interference from background noise can be avoided, improving the reliability of the search.
[0063] S312. Based on the amplitude of each frequency point in the current local neighborhood, the mean of the amplitude in the current local neighborhood, and the standard deviation of the amplitude in the current local neighborhood, obtain the statistical significance index of each frequency point in the current local neighborhood. In this embodiment, the statistical significance index is defined as: ; in, The amplitude at frequency point f; It is the mean of the amplitude in the local neighborhood of the frequency point f; Let f be the standard deviation of the amplitude within the local neighborhood of the frequency point f; The statistical significance index of frequency point f is... It reflects the significance of the spectral line amplitude at frequency point f relative to its local background.
[0064] In this embodiment, for each frequency point within the current local neighborhood, a statistical significance index is calculated. This index is defined as the difference between the amplitude of that point and the mean amplitude of the neighborhood, divided by the standard deviation of the amplitudes within the neighborhood. The larger the index value, the more prominent the frequency point is relative to its local background. This standardization process makes spectral regions with different amplitude levels comparable without relying on a global threshold.
[0065] S313. When the maximum value of the statistical significance index of each local maximum point in the current local neighborhood is less than the preset threshold, and the statistical significance index of each of the preset number of consecutive frequency points in the search direction is less than the preset proportion of the preset threshold, stop the search and determine the current frequency as the boundary in the current search direction.
[0066] In this embodiment, the condition for stopping the search is: and N frequency points appear consecutively; in, It is the set of statistical significance indicators for all local maxima within the current local neighborhood; The preset threshold is N (usually 2-3); N is the number of points required for continuity (usually 3-5); when and When both of these conditions are met, the current frequency is determined as the boundary in the current search direction.
[0067] In this embodiment, a preset threshold is used to determine whether a significant frequency shifting component exists. When the maximum value of the statistical significance index of all local maxima within a local neighborhood is still lower than the threshold, it indicates that there are no significant prominent frequencies within that neighborhood. Simultaneously, to confirm that the frequency shifting component has indeed disappeared rather than experiencing an occasional decrease, it is also required that the statistical significance index of multiple consecutive frequency points along the search direction is lower than a certain low percentage (e.g., half) of the threshold. When both conditions are met simultaneously, the frequency shifting fluctuation is considered to have reached a boundary, the search is stopped, and the current frequency point is determined as either a low-frequency boundary or a high-frequency boundary. This dual-judgment mechanism effectively avoids misjudging the boundary due to a single point's occasional decrease, effectively improving the robustness of boundary determination.
[0068] The search process in the high-frequency direction is completely symmetrical to that in the low-frequency direction, that is, from... The process begins by moving towards higher frequencies, using the same local neighborhood definition, local maxima identification, statistical significance index calculation, and termination conditions, until the high-frequency boundary is determined. Through this bilateral iterative search, the complete low-frequency and high-frequency boundaries of the frequency shift fluctuation can be obtained.
[0069] It should be noted that the core of the rotational speed fluctuation identification method provided in this application lies in using frequency domain statistical baselines to identify stable frequency components in rotating machinery that are related to a reference frequency (such as the rotational frequency) in an anti-interference manner. Those skilled in the art will understand that this method is not limited to the identification and tracking of the rotational frequency (fundamental frequency) itself. Based on the same technical principle, by adjusting the reference frequency value (e.g., setting the reference frequency to 2 times, 3 times, etc. of the rotational frequency), or by performing demodulation analysis on the spectrum, this method is also applicable to identifying and tracking all components with a deterministic frequency relationship to the rotational frequency, such as the harmonics of the rotational frequency and the modulation sidebands generated by modulation phenomena. This helps to more comprehensively analyze the fault characteristics of the equipment, such as the modulation sidebands of gear meshing frequencies (harmonics of the rotational frequency) or bearing fault characteristic frequencies.
[0070] like Figure 2 As shown in the figure, an embodiment of this application discloses a speed fluctuation identification system, the system comprising: Acquisition module 21 is used to acquire the spectrum of vibration signals from rotating machinery; Sorting module 22 is used to sort the amplitude values in the spectrum to obtain an ordered spectrum; The positioning module 23 is used to determine, in the ordered spectrum, a percentile window defined by the first high percentile and the second high percentile below the first high percentile, and to locate the amplitude range corresponding to the percentile window; Processing module 24 is used to obtain statistical baseline values based on the various amplitudes within the amplitude range; The identification module 25 is used to identify the frequency of vibration signals using statistical baseline values.
[0071] The above description is merely an embodiment of the present invention. It should be noted that those skilled in the art can make improvements without departing from the inventive concept of the present invention, but these improvements all fall within the protection scope of the present invention.
Claims
1. A method for identifying speed fluctuations, characterized in that, The method includes: Obtain the spectrum of vibration signals from rotating machinery; The amplitude values in the spectrum are sorted to obtain an ordered spectrum; In the ordered spectrum, a percentile window is determined by a first high percentile and a second high percentile below the first high percentile, and the amplitude range corresponding to the percentile window is located. Based on each amplitude within the stated amplitude range, a statistical baseline value is obtained; The frequency of the vibration signal is identified using the statistical baseline value.
2. The method according to claim 1, characterized in that, The first high percentile is 85% to 95%, and the second high percentile is 10% to 20% lower than the first high percentile.
3. The method according to claim 1, characterized in that, The step of obtaining the statistical baseline value based on each amplitude within the amplitude range includes: A weighted average based on the location index is performed on each amplitude within the amplitude range to obtain a weighted average result, and the weighted average result is used as the statistical baseline value.
4. The method according to claim 3, characterized in that, The weights used in the weighted average are calculated by applying a Gaussian function decay based on the distance between the position index and the center position of the amplitude range, or by applying a linear decay based on the distance between the position index and the center position of the amplitude range.
5. The method according to claim 1, characterized in that, The step of identifying the frequency conversion of the vibration signal using the statistical baseline value includes: Obtain the reference frequency; Spectral peaks are searched within the neighborhood of the reference frequency, and the confidence level of each spectral peak is obtained using the statistical baseline value, the amplitude of each spectral peak, the standard deviation of the spectral amplitude within the neighborhood of the reference frequency, the frequency of the spectral peak, the reference frequency, and the frequency offset penalty coefficient. The frequency corresponding to the spectral peak of the maximum value among the various confidence levels is determined as the initially identified frequency.
6. The method according to claim 5, characterized in that, The confidence level is determined by both the amplitude difference factor and the frequency offset factor. The amplitude difference factor is used to assess the significance of the amplitude of the spectral line peak based on the deviation between the amplitude of the spectral line peak and the statistical baseline value, as well as the standard deviation of the spectral line amplitude in the neighborhood of the reference frequency. The frequency offset factor is used to evaluate the frequency matching degree of the spectral peak based on the deviation between the frequency of the spectral peak and the reference frequency, as well as the frequency offset penalty coefficient.
7. The method according to claim 5, characterized in that, Also includes: The reliability index of the initially identified frequency is obtained based on the amplitude of the spectral line corresponding to the reference frequency, the statistical baseline value, the maximum amplitude of the spectral line in the neighborhood of the reference frequency, the variance of the spectral line amplitude in the local neighborhood centered on the initially identified frequency, and the variance of the spectral line amplitude in the neighborhood of the reference frequency. The reliability index is used to evaluate the reliability of the initially identified frequency.
8. The method according to claim 5, characterized in that, Also includes: Using the initially identified frequency as the center frequency, a bilateral iterative search is performed in both the low-frequency and high-frequency directions to determine the low-frequency and high-frequency boundaries of the frequency fluctuation.
9. The method according to claim 8, characterized in that, The bilateral iterative search includes: In each iteration, determine the local maximum point in the current local neighborhood of the current frequency point; Based on the amplitude of each frequency point in the current local neighborhood, the mean amplitude in the current local neighborhood, and the standard deviation of the amplitude in the current local neighborhood, the statistical significance index of each frequency point in the current local neighborhood is obtained. When the maximum value of the statistical significance index of each local maximum point in the current local neighborhood is less than a preset threshold, and the statistical significance index of each of the preset number of consecutive frequency points in the search direction is less than a preset proportion of the preset threshold, the search is stopped, and the current frequency is determined as the boundary in the current search direction.
10. A speed fluctuation identification system, characterized in that, The system includes: The acquisition module is used to acquire the spectrum of vibration signals from rotating machinery. The sorting module is used to sort the amplitude values in the spectrum to obtain an ordered spectrum; The positioning module is used to determine, in the ordered spectrum, a percentile window defined by a first high percentile and a second high percentile lower than the first high percentile, and to locate the amplitude range corresponding to the percentile window; The processing module is used to obtain a statistical baseline value based on each amplitude within the amplitude range; The identification module is used to identify the frequency of the vibration signal using the statistical baseline value.