A method for extracting the failure gear rotation frequency and meshing frequency of a gear box

Abstract

The application discloses a kind of gear box fault gear rotation frequency and meshing frequency extraction method, belong to mechanical equipment condition monitoring and fault diagnosis technical field.The method is by collecting gear box vibration acceleration signal, respectively generate spectrum sequence and frequency domain autocorrelation spectrum sequence;After spectrum flattening, strong peak tailing, noise scale estimation, adaptive threshold peak extraction, frequency correction, redundant peak elimination, frequency family consistency score, candidate meshing frequency is screened from spectrum sequence, and candidate fault gear rotation frequency is screened from frequency domain autocorrelation spectrum sequence in combination with rotation frequency interval.The application does not need to rely on gear tooth number, transmission ratio and other structural prior information, can automatically suppress noise and resonance interference, accurately extract fault characteristic frequency, applicable to engineering field without equipment data, weak early fault diagnosis scene, provide reliable basis for gear box fault early warning and positioning.

Description

Technical Field

[0001] This invention belongs to the field of mechanical equipment condition monitoring and fault diagnosis technology, specifically relating to a method for extracting the rotational frequency and meshing frequency of faulty gears in a gearbox. Background Technology

[0002] Gearboxes are widely used in automobiles, wind power, mining machinery, aerospace, and high-end equipment. Their operating status directly affects the overall transmission efficiency, service reliability, and operational safety. Early-stage localized gear damage typically manifests as weak impacts, weak modulation, and narrow-band structural changes, often appearing in the frequency domain as the gradual generation and enhancement of sidebands near the meshing frequency. Therefore, accurately extracting the fault-related gear rotation frequencies and meshing frequencies is crucial for early gearbox diagnosis, fault location, and condition assessment.

[0003] Existing methods for extracting fault frequencies mostly rely on the number of teeth, transmission ratio, shaft mapping relationship, speed measurement results, or human experience prediction. In scenarios with complete structural priors, these methods can usually achieve good results; however, in engineering field testing, situations where historical structural information is missing, parts are frequently replaced, multiple models are used interchangeably, or testing conditions are limited, it is often difficult to obtain complete and reliable structural priors, making it difficult for traditional methods to directly obtain the rotational frequency and meshing frequency of the faulty gear.

[0004] Meanwhile, if an exhaustive search is performed across the entire frequency band, the number of observable spectral lines will increase rapidly with the improvement of frequency resolution and the expansion of observation bandwidth, resulting in high computational costs. At the same time, although gear fault modulation sidebands have a typical structure of equal intervals around the center frequency, in actual engineering signals they are often affected by structural resonance background fluctuations, spectral leakage, rotational speed fluctuations, broadband noise, and occasional narrowband interference, causing the sideband positions and amplitudes to shift and distort relative to theoretical values. Relying solely on fixed thresholds, local manual frequency selection, or single envelope spectrum analysis can easily lead to problems such as missed detection of weak physical frequencies, misselection of strong interference peaks, and excessive threshold sensitivity. Summary of the Invention

[0005] The technical problem this invention aims to solve is to address the shortcomings of the prior art by providing a method for extracting the gear rotation frequency and meshing frequency in gearbox faults. Through steps such as spectral flattening, adaptive threshold significant peak extraction, redundant peak removal, harmonic family consistency scoring, and candidate sideband group combination discrimination, this method effectively suppresses the influence of noise, resonance background fluctuations, and narrowband interference on the frequency identification results, achieving accurate identification of gearbox fault characteristic frequencies under unstructured prior conditions. This solves the problems of insufficient stability and weak anti-interference ability in the extraction of gear fault-related frequencies under unstructured prior conditions in existing technologies.

[0006] This invention adopts the following technical solution: a method for extracting the rotational frequency and meshing frequency of a faulty gear in a gearbox, comprising the following steps: The vibration acceleration signal of the gearbox is collected, the spectrum of the signal is estimated to obtain the spectrum sequence, and the frequency domain autocorrelation operation is performed on the spectrum sequence to obtain the frequency domain autocorrelation spectrum sequence; Perform spectral flattening on the spectral sequence and the frequency domain autocorrelation spectral sequence respectively; Strong peak truncation is performed on the flattened sequence; Based on the truncated sequence, the median absolute error method is used to estimate the noise scale and generate an adaptive significant peak detection threshold. Candidate peaks are extracted based on the threshold, and the frequency accuracy of the candidate peaks is corrected. Remove the overtones and redundant peaks of similar frequencies from the candidate peaks to obtain a purified candidate peak set; The candidate frequencies in the purification candidate peak set are scored and sorted according to the frequency family consistency score to obtain the potential physical frequency sequence. The set of candidate faulty gear meshing frequencies is selected from the potential physical frequencies of the spectral sequence, and the set of candidate faulty gear rotation frequencies is selected from the potential physical frequencies of the frequency domain autocorrelation spectral sequence in combination with the rotation frequency search interval.

[0007] Preferably, the spectral flattening process includes estimating the spectral background using a wide window mid-range filter, and then reducing the influence of the background amplitude on peak detection by ratio flattening.

[0008] Preferably, the number of spectral lines in the wide window mid-range filter is determined based on the sampling frequency and frequency resolution.

[0009] Preferably, the strong peak truncation process includes performing upper quantile truncation on the flattened sequence to remove strong peaks that exceed the upper truncation limit, thus forming a truncated sequence.

[0010] Preferably, the noise scale estimation is to calculate the median of the truncated sequence, calculate the median absolute error based on the median, and convert the median absolute error into an equivalent noise standard deviation.

[0011] Preferably, the adaptive significant peak detection threshold is obtained by multiplying the equivalent noise standard deviation by a preset significance factor and adding the median.

[0012] Preferably, the candidate peak is a spectral peak whose amplitude exceeds the adaptive significant peak detection threshold and is a local maximum; the frequency accuracy correction uses a three-point local parabolic interpolation method to correct the true frequency position of the candidate peak.

[0013] Preferably, the redundant peak removal includes determining the harmonics and similar frequency relationships between any two candidate frequencies, retaining the fundamental frequency peak with the smaller frequency value, and removing the corresponding harmonic peaks and similar interference peaks.

[0014] Preferably, the harmonic family consistency score comprehensively considers the consistency of the number and amplitude of harmonic peaks of candidate frequencies, thereby reducing the dominant influence of higher-order harmonics on the scoring results.

[0015] Preferably, the frequency search range is set according to the actual rotational speed range of the equipment, and the candidate meshing frequency and the candidate faulty gear frequency are both selected from the preset number of potential physical frequencies with the highest scores.

[0016] Compared with the prior art, the present invention has at least the following beneficial effects: By collecting gearbox vibration acceleration signals and performing spectral analysis and frequency domain autocorrelation analysis, candidate meshing frequency sets and candidate rotation frequency sets are constructed through spectral flattening, adaptive threshold significant peak extraction, redundant peak removal, and harmonic family consistency scoring. By combining candidate sideband group combination discrimination and consistency scoring, fault-related gear rotation frequencies and their corresponding meshing frequencies are automatically extracted, realizing accurate identification of gearbox fault characteristic frequencies under unstructured prior conditions, and providing a basis for early fault warning and fault location.

[0017] In summary, the method for extracting the rotational frequency and meshing frequency of faulty gears in a gearbox according to the present invention can achieve joint extraction of fault-related gear rotational frequency and meshing frequency by utilizing the spectral structure information of the vibration signal itself, without relying on the number of teeth, transmission ratio, shaft system mapping relationship, or expert experience, even without structural prior knowledge. This significantly improves the applicability and accuracy of fault frequency extraction under conditions of lack of prior knowledge.

[0018] By employing steps such as spectral flattening, adaptive threshold significant peak extraction, redundant peak removal, harmonic family consistency scoring, and candidate sideband group combination discrimination, the method effectively suppresses the impact of noise, resonance background fluctuations, and narrowband interference on frequency identification results, thus improving its robustness and stability under complex operating conditions. Even for signals with weak sidebands and indistinct features in the early stages of a fault, this invention can still effectively extract weak fault symptoms. Furthermore, this invention provides reliable frequency information for subsequent fault location, condition assessment, and predictive maintenance, demonstrating significant engineering application value.

[0019] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the following description of the relative embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the gearbox structure in an embodiment; Figure 3 The following is a time-domain waveform diagram of the vibration acceleration signal for an example. Figure 4 This is an example of the application effect in the early fault spectrum and frequency domain autocorrelation spectrum of the embodiment; Figure 5 The example shows a score graph of the consistency of the frequency doubling family of the spectral candidate frequencies. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "one side," "one end," and "one side," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0024] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0025] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0026] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0027] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0028] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0029] like Figure 1 As shown, a method for extracting the rotational frequency and meshing frequency of a faulty gear in a gearbox includes the following steps: S1. Collect the vibration acceleration signal of the gearbox, perform spectrum estimation on the signal to obtain the spectrum sequence, and perform frequency domain autocorrelation operation on the spectrum sequence to obtain the frequency domain autocorrelation spectrum sequence.

[0030] Specifically, the vibration acceleration signal of the gearbox is subjected to spectral estimation to obtain a spectral sequence. Simultaneously, frequency domain autocorrelation is performed on the spectral sequence to obtain a frequency domain autocorrelation spectral sequence. The spectral sequence is mainly used to extract candidate meshing frequencies, while the frequency domain autocorrelation spectral sequence is mainly used to extract candidate faulty gear rotation frequencies.

[0031] S2. Perform spectral flattening on the spectral sequence and the frequency domain autocorrelation spectral sequence respectively.

[0032] Spectral flattening involves estimating the spectral background using a wide-window mid-range filter, and then reducing the impact of background amplitude on peak detection through ratio flattening. Specifically, it includes the following: S21. Integrate the extraction of the two frequencies using a unified framework. Provide a spectral amplitude sequence corresponding to equally spaced frequencies. spectral amplitude sequence The corresponding frequency sequence is The frequency resolution is The spectral sequence can be represented as: ; in, This indicates the background trend or structural resonance envelope; It represents narrowband spectral line components, including meshing frequencies, side frequencies, local resonance peaks, etc.

[0033] S22. Perform background estimation and spectral flattening to avoid background frequency components raising the peak detection threshold.

[0034] The spectral background is estimated using a wide-window mid-range filtering method. ; in, It is the number of spectral lines contained in the filter window.

[0035] The number of spectral lines in a wide-window mid-range filter is determined based on the sampling frequency and frequency resolution. Specifically, the empirical formula for determining its value is as follows: .in Sampling frequency, This refers to the frequency resolution. This is a preset multiple, usually between 5 and 10.

[0036] A ratio flattening method is used to flatten the spectral sequence to reduce the influence of background on the peak detection threshold. ; in, ε is the amplitude of the k-th spectral line after flattening; ε is a very small positive number.

[0037] S3. Perform strong peak truncation on the flattened sequence.

[0038] Strong peak truncation involves truncating the flattened sequence at the upper quantile, removing strong peaks that exceed the upper truncation limit to form a truncated sequence. Specifically, it includes the following: Since strong meshing frequency peaks or structural resonance peaks significantly amplify noise scale estimation, to improve the robustness of the method, strong peaks are treated as outliers when calculating the peak detection threshold, thus avoiding their participation in noise scale estimation.

[0039] For flattened sequences Perform upper quantile truncation. Let the truncation quantile be . Commonly used Obtain the upper bound of the truncation. And define the truncated sequence as:

[0040] The purpose of introducing truncation is to treat extremely strong meshing peaks and occasional interference peaks as outliers, thereby avoiding their dominance in noise scaling estimation.

[0041] In this embodiment, based on the results obtained from a large number of spectral sequence experiments A good balance is achieved between suppressing strong peaks and maintaining the statistical stability of noise.

[0042] S4. Based on the truncated sequence, the median absolute error method is used to estimate the noise scale and generate an adaptive significant peak detection threshold.

[0043] Noise scaling involves calculating the median of the truncated sequence, calculating the median absolute error based on the median, and then converting the median absolute error into an equivalent noise standard deviation. Specifically, this includes the following: Based on the obtained truncated sequences, a MAD (median absolute deviation) noise scaling method is proposed. Calculate the median using the following formula. And the median absolute error (MAD): ; ; Gearbox vibration noise can be approximated as Gaussian noise, and its equivalent standard deviation can be estimated using MAD. .

[0044] Based on the noise estimation scale, the adaptive significant peak detection threshold is obtained by multiplying the equivalent noise standard deviation by a preset significance factor and adding the median.

[0045] Specifically, adaptive saliency peak detection threshold .in, Given a significance factor, typically ranging from 3 to 6, the probability of a single false alarm increases with... Increases and decreases.

[0046] S5. Extract candidate peaks based on the threshold and correct the frequency accuracy of the candidate peaks.

[0047] Candidate peaks are spectral peaks whose amplitude exceeds the adaptive significant peak detection threshold and are local maxima. Frequency accuracy correction uses a three-point local parabolic interpolation method to correct the true frequency position of the candidate peaks. Details are as follows: If the peak exceeds the significant peak threshold and is a local maximum, then the peak is considered a candidate peak, and the determination criterion is as follows:

[0048] Since spectral peaks often fall between discrete frequency grids, to improve the accuracy of peak frequency estimation, each candidate peak... The offset is obtained by using three-point local parabolic interpolation: ; The true location of the significant peak can be obtained. .

[0049] The significant peak amplitude after interpolation of the three-point local parabolic curve is: After interpolation, the corrected and accurate set of significant peak frequencies can be obtained. .

[0050] S6. Remove the overtones and redundant peaks of similar frequencies from the candidate peaks to obtain a set of purified candidate peaks.

[0051] Redundant peak removal involves determining the harmonics and similar frequency relationships between any two candidate frequencies, retaining the fundamental frequency peak with the smaller frequency value, and removing the corresponding harmonic peaks and similar interference peaks. The specific steps are as follows: For any two candidate frequencies First, estimate its nearest integer multiple using the following formula: ; The following formula defines the consistency error of harmonics and similar frequencies: ; If the judgment condition is met: Then it is believed for The frequency of the interference is a multiple or similar to the frequency of the interference.

[0052] in, The absolute deviation threshold; This is the relative deviation threshold.

[0053] In this case, only the peak with the lower frequency is retained, and the other frequency is removed from the candidate set. The final candidate peak set is obtained after removing harmonics and similar frequencies. This refers to the set of purification candidate peaks and the corresponding set of peak amplitudes: .

[0054] S7. Perform frequency harmonic consistency scoring and sorting on the candidate frequencies in the purification candidate peak set to obtain the potential physical frequency sequence.

[0055] The harmonic family consistency score comprehensively considers the consistency of the number and amplitude of harmonic peaks at candidate frequencies, reducing the dominant influence of higher-order harmonics on the scoring results. Specifically, it includes the following: For any candidate frequency Construct its frequency doubling family Determine the target frequency. Does a corresponding peak exist? Then it is considered that the frequency multiplication is It was observed in the candidate set, and the amplitude of the octave peak was denoted as... .

[0056] To simultaneously reflect both the number and intensity of octave peaks, and to avoid excessive dominance of higher-order octaves in the scoring, the octave family consistency score is defined as follows: ; If a candidate frequency is a true characteristic frequency, its harmonic chain is more complete and its overall amplitude is more consistent, thus obtaining a higher score. For all candidate frequencies... ,calculate And sorted by rating from highest to lowest. The final output will be the one with the highest score. The candidate frequencies are given as a set of potential physical frequencies, and a sequence of potential physical frequencies sorted by score is provided. .

[0057] S8. Select the set of candidate faulty gear meshing frequencies from the potential physical frequencies of the spectrum sequence, and select the set of candidate faulty gear rotation frequencies from the potential physical frequencies of the frequency domain autocorrelation spectrum sequence in combination with the rotation frequency search interval.

[0058] Both the candidate meshing frequency and the candidate faulty gear rotation frequency are selected from the preset number of potential physical frequencies with the highest scores. Details are as follows: For the spectral sequence, by retaining the high-scoring frequencies as potential meshing frequencies and their harmonics according to the above sorting, a set of candidate meshing frequencies can be obtained: For the autocorrelation spectrum sequence, frequencies with high scores are retained as potential meshing frequencies and their harmonics according to the above sorting, and necessary physical constraints are introduced to search only within a given frequency transition interval. The candidate peak values ​​are retained, and this range can be set based on the equipment's speed range or common sense. This yields a set of candidate faulty gear frequencies. .

[0059] In this embodiment, taking the gear accelerated life test as an example, the rotational frequency and meshing frequency of the faulty gear in the gearbox are extracted, as follows: A schematic diagram of the gearbox structure is shown below. Figure 2 As shown, the acquired signal is the vibration acceleration signal of the test gearbox, which includes 6 channels, specifically input axes X, Y, and Z, and output axes X, Y, and Z. The sampling frequency is 12.8kHz, the sampling duration is 2.56s, and the sampling interval is 2min. The input rotation frequency is 40Hz. The meshing frequencies and the rotation frequencies of each shaft are shown in Table 1 below.

[0060] Table 1. Rotational frequency and meshing frequency of each gear in the gearbox

[0061] The equipment's lifespan was set at 110 hours. The actual failure mode was the continuous breakage of four teeth on gear III, with pitting on all tooth surfaces of gears III and IV. Because the faulty gear was far from the input shaft, the fault signal was weak. To verify the effectiveness of this method in extracting weak fault features, this paper used the vibration acceleration signal in the Y direction of the input shaft (this direction is the loading direction).

[0062] In this experiment, the full-life vibration time-domain waveform of the gearbox was obtained as follows: Figure 3 As shown in the figure, the amplitude of the vibration signal in the time domain begins to increase at 3080 minutes, which is considered to be the early fault moment. The data at this moment is analyzed to extract the rotational frequency and meshing frequency of the faulty gear.

[0063] like Figure 4 As shown, in the spectral sequence branch, after spectral flattening and adaptive threshold discrimination, the candidate spectral lines exceeding the detection threshold are mainly composed of two sets of meshing frequency components and some harmonics, indicating that this branch can effectively preserve the main physical frequency information related to gear transmission in a complex spectral background.

[0064] At the same time, only a small number of narrowband interference lines with large amplitudes were retained, indicating that after the aforementioned processing, random noise and a large number of non-critical stray lines have been well suppressed, thus enabling the candidate frequency set to have good compactness while maintaining key structural information.

[0065] Secondly, in the frequency domain autocorrelation spectrum branch, only the candidate interval frequencies corresponding to the shaft rotation frequency of the faulty gear are retained. This indicates that even when the sidebands are not very significant in the early stage of the fault and individual sideband components are difficult to identify directly, the frequency domain autocorrelation spectrum can still effectively strengthen the equidistant repetition relationship between spectral lines, thereby highlighting the frequency candidates related to the modulation interval.

[0066] like Figure 5 As shown, based on this, after the harmonic family consistency score, the meshing frequency component that is consistent with the actual transmission structure received a higher score, while the scores of isolated interference spectral lines that, although high in amplitude, lack complete harmonic chain support were significantly suppressed.

[0067] The results show that the harmonic family consistency score can not only further improve the identification priority of the real physical frequency, but also effectively weaken the adverse effects of strong interference spectral lines on the subsequent candidate frequency ranking and screening results, thus providing a more reliable basis for the accurate extraction of fault-related meshing frequencies and rotation frequencies.

[0068] In summary, this invention provides a method for extracting the gear rotation frequency and meshing frequency in gearbox faults. It requires no structural information such as the number of teeth, transmission ratio, or shaft system relationship, and also requires no manual experience or prior knowledge, making it suitable for various complex engineering scenarios. Through multiple processing methods including spectral flattening, peak truncation, adaptive thresholding, and octave scoring, it effectively suppresses noise, resonance, and speed fluctuation interference, exhibiting strong anti-interference capabilities. Frequency domain autocorrelation enhances sideband spacing characteristics, accurately identifying the characteristic frequencies of early, weak faults. It possesses weak fault extraction capabilities, requiring no manual intervention throughout the process, avoiding errors from manual frequency selection and threshold setting, and providing stable and reliable extraction results. It can be directly used for gearbox fault early warning, location, and predictive maintenance, possessing high practical value.

[0069] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A method for extracting the rotational frequency and meshing frequency of a faulty gear in a gearbox, characterized in that, Includes the following steps: The vibration acceleration signal of the gearbox is collected, the spectrum of the signal is estimated to obtain the spectrum sequence, and the frequency domain autocorrelation operation is performed on the spectrum sequence to obtain the frequency domain autocorrelation spectrum sequence; Perform spectral flattening on the spectral sequence and the frequency domain autocorrelation spectral sequence respectively; Strong peak truncation is performed on the flattened sequence; Based on the truncated sequence, the median absolute error method is used to estimate the noise scale and generate an adaptive significant peak detection threshold. Candidate peaks are extracted based on the threshold, and the frequency accuracy of the candidate peaks is corrected. Remove the overtones and redundant peaks of similar frequencies from the candidate peaks to obtain a purified candidate peak set; The candidate frequencies in the purification candidate peak set are scored and sorted according to the frequency family consistency score to obtain the potential physical frequency sequence. The set of candidate faulty gear meshing frequencies is selected from the potential physical frequencies of the spectral sequence, and the set of candidate faulty gear rotation frequencies is selected from the potential physical frequencies of the frequency domain autocorrelation spectral sequence in combination with the rotation frequency search interval.

2. The method for extracting the rotational frequency and meshing frequency of faulty gears in a gearbox according to claim 1, characterized in that, The spectral flattening process includes estimating the spectral background using a wide window mid-range filter, and then reducing the impact of the background amplitude on peak detection by ratio flattening.

3. The method for extracting the rotational frequency and meshing frequency of faulty gears in a gearbox according to claim 2, characterized in that, The number of spectral lines in the wide window mid-range filter is determined based on the sampling frequency and frequency resolution.

4. The method for extracting the rotational frequency and meshing frequency of faulty gears in a gearbox according to claim 1, characterized in that, The strong peak truncation process includes truncating the flattened sequence at the upper quantile, removing strong peaks that exceed the upper truncation limit, and forming a truncated sequence.

5. The method for extracting the rotational frequency and meshing frequency of faulty gears in a gearbox according to claim 1, characterized in that, The noise scale estimation involves calculating the median of the truncated sequence, calculating the median absolute error based on the median, and converting the median absolute error into an equivalent noise standard deviation.

6. The method for extracting the rotational frequency and meshing frequency of faulty gears in a gearbox according to claim 5, characterized in that, The adaptive significant peak detection threshold is obtained by multiplying the equivalent noise standard deviation by a preset significance factor and adding the median.

7. The method for extracting the rotational frequency and meshing frequency of faulty gears in a gearbox according to claim 1, characterized in that, The candidate peak is a spectral peak whose amplitude exceeds the adaptive significant peak detection threshold and is a local maximum; the frequency accuracy correction uses a three-point local parabolic interpolation method to correct the true frequency position of the candidate peak.

8. The method for extracting the rotational frequency and meshing frequency of faulty gears in a gearbox according to claim 1, characterized in that, The redundant peak removal includes determining the harmonics and similar frequency relationships between any two candidate frequencies, retaining the fundamental frequency peak with the smaller frequency value, and removing the corresponding harmonic peaks and similar interference peaks.

9. The method for extracting the rotational frequency and meshing frequency of a faulty gear in a gearbox according to claim 1, characterized in that, The harmonic family consistency score comprehensively considers the consistency of the number and amplitude of harmonic peaks of candidate frequencies, reducing the dominant influence of higher-order harmonics on the scoring results.

10. The method for extracting the rotational frequency and meshing frequency of a faulty gear in a gearbox according to claim 1, characterized in that, The frequency search range is set according to the actual speed range of the equipment, and the candidate meshing frequency and the candidate faulty gear frequency are both selected from the preset number of potential physical frequencies with the highest scores.

Images