A Mechanical Transient Fault Diagnosis Method and Device Based on Group Delay Ridge Optimization

Abstract

The present invention discloses a mechanical transient fault diagnosis method and device based on group delay ridge optimization, belonging to the technical field of intelligent mechanical fault diagnosis, and is used to solve the technical problems that the existing vibration signal processing methods are difficult to extract the transient signal features from the vibration signals with complex components, are easily affected by interference, and have a low fault diagnosis accuracy. The method includes: performing time-frequency transformation and phase modulation wavelet transformation on the vibration signal to obtain the corresponding time-frequency transformation result; extracting the group delay ridge of the first mode in the vibration signal; performing group delay band filtering on the group delay ridge to obtain the optimized group delay ridge; removing the time-frequency coefficients corresponding to the optimized group delay ridge, iteratively extracting the group delay ridge of the second mode, and optimizing the ridge line of the searched group delay ridge until the group delay ridge of each mode is obtained; obtaining the optimized time-frequency transformation result according to the group delay ridge of each mode, and performing mechanical transient fault diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent diagnosis of mechanical faults, and particularly to a mechanical transient fault diagnosis method and device based on the optimization of group delay ridge lines. Background Art

[0002] Mechanical signal analysis technology is widely used in various fields such as aerospace, vehicles, and manufacturing. Since industrial equipment often operates continuously under heavy loads, fatigue and other complex working conditions for a long time, its core components and key structures will inevitably suffer varying degrees of damage. When a rotating component in a mechanical device fails, it often appears in the monitored vibration signal in the form of a transient impact, such as local wear, rotor misalignment, surface corrosion and other problems. By processing and analyzing such signals, it helps to detect signs of faults in advance, identify the types and locations of faults in the mechanical system, and thus provide a strong basis for timely and effective maintenance.

[0003] In engineering practice, the operating state of rotating mechanical equipment can be measured and analyzed through various physical signals such as vibration, sound, pressure, and temperature. Vibration signals are easy to collect, sensitive to machine faults, and can separate the parts of interest from other strong vibration components that are not needed. Therefore, vibration signals are very useful in detecting early defects, extracting fault characteristics, optimizing health indicators, etc., which makes the fault diagnosis technology based on vibration signals widely studied and applied in the state detection of mechanical equipment. When industrial equipment fails, its vibration often shows complex modulation change rules, which in turn causes a sharp change in the derivative of the phase function. Due to the complex mechanical structure and harsh working environment, the components of the monitored signal generated are becoming more complex, resulting in the difficulty of traditional signal processing methods to meet the requirements of the above transient signal feature extraction and fault identification, especially for early mechanical faults. The diagnostic signals are often affected by various interferences from the environment, sensors, and the mechanical system itself, and the impact characteristics caused may be submerged by strong background noise, resulting in misdiagnosis. Summary of the Invention

[0004] Embodiments of the present invention provide a mechanical transient fault diagnosis method and device based on the optimization of group delay ridge lines, which are used to solve the following technical problems: The existing vibration signal processing methods are difficult to extract transient signal features from complex vibration signals, are easily affected by interference, and have a low fault diagnosis accuracy rate.

[0005] Embodiments of the present invention adopt the following technical solutions:

[0006] On the one hand, embodiments of the present invention provide a mechanical transient fault diagnosis method based on the optimization of group delay ridge lines. The method includes: performing time-frequency transformation and phase modulation wavelet transformation on the collected vibration signal to obtain the corresponding time-frequency transformation result;

[0007] Based on the time-frequency transformation result, extract the group delay ridge line of the first mode in the vibration signal;

[0008] Perform group delay band filtering on the group delay ridge line to remove noise interference and obtain an optimized group delay ridge line;

[0009] After removing the time-frequency coefficients corresponding to the optimized group delay ridge line, iteratively extract the group delay ridge line of the second mode, and optimize the ridge line of the searched group delay ridge line; and so on until the group delay ridge lines of each mode are obtained;

[0010] According to the group delay ridge lines of each mode, obtain an optimized time-frequency transformation result, and perform time-frequency compression on the optimized time-frequency transformation result to obtain an optimized time-frequency image corresponding to the vibration signal;

[0011] Diagnose the transient fault type of the current machine based on the group delay ridge lines of each mode.

[0012] In a feasible implementation manner, perform time-frequency transformation and phase-modulated wavelet transformation on the collected vibration signal to obtain a corresponding time-frequency transformation result, specifically including:

[0013] Model the collected vibration signal as a superposition of several intrinsic mode functions, and perform Fourier transform to obtain a time-frequency signal model;

[0014] Perform phase modulation on the wavelet basis function through a scale factor and a translation factor, and perform wavelet transform on the time-frequency signal model through the modulated wavelet basis function to obtain a corresponding time-frequency transformation result.

[0015] In a feasible implementation manner, based on the time-frequency transformation result, extract the group delay ridge line of the first mode in the vibration signal, specifically including:

[0016] Divide the time-frequency plane where the time-frequency transformation result is located into several time-frequency segments along the frequency direction;

[0017] Determine the point with the maximum time-frequency energy in each time-frequency segment as the search starting point, and sequentially search for the time-frequency energy peak points in the neighborhood to the left and right; connect the point with the maximum time-frequency energy in the time-frequency segment with the time-frequency energy peak points to obtain the group delay ridge line of each time-frequency segment;

[0018] Obtain the time-frequency energy of each time-frequency segment, and determine the time-frequency segment with the maximum time-frequency energy as the first mode; determine the group delay ridge line corresponding to the time-frequency segment with the maximum time-frequency energy as the group delay ridge line of the first mode.

[0019] In a feasible implementation, group delay band filtering is performed on the group delay ridge line to remove noise interference and obtain an optimized group delay ridge line, which specifically includes:

[0020] Taking the time-frequency energy peak of the time-frequency transformation result as the group delay time-width center, and setting the group delay ridge line at the group delay time-width center to align with the reference value;

[0021] Based on the alignment reference value, re-adjust the definition of the effective time-width region in the group delay estimation process;

[0022] Based on the definition of the effective time-width region, determine the group delay estimator;

[0023] Calculate the group delay band width of each group delay ridge line in the group delay estimator, count the number of each group delay band width, and draw a distribution histogram of the group delay band width;

[0024] Analyze the distribution histogram to obtain the filtering threshold;

[0025] Based on the filtering threshold, perform filtering on the group delay bands in the group delay ridge line to remove the noise interference in the group delay ridge line and obtain an optimized group delay ridge line.

[0026] In a feasible implementation, analyzing the distribution histogram to obtain the filtering threshold specifically includes:

[0027] Fitting the distribution histogram by the Trust-Region algorithm to analyze the distribution properties of the noise signal and the normal signal;

[0028] Based on the distribution properties, obtain the intersection point of the noise signal distribution region and the normal signal distribution region, and use the group delay band width corresponding to the intersection point as the filtering threshold.

[0029] In a feasible implementation, based on the filtering threshold, performing filtering on the group delay bands in the group delay ridge line specifically includes:

[0030] Filter out the group delay bands in the group delay ridge line with a group delay band width less than the filtering threshold, and retain the group delay bands with a width greater than or equal to the filtering threshold to obtain a filtered group delay ridge line.

[0031] In a feasible implementation, after removing the time-frequency coefficients corresponding to the optimized group delay ridge line, iteratively extract the group delay ridge line of the second mode, and perform ridge line optimization on the searched group delay ridge line; and so on until the group delay ridge lines of each mode are obtained, which specifically includes:

[0032] Set the time-frequency coefficients corresponding to the group delay ridge line of the optimized first mode to zero in the time-frequency transformation result;

[0033] Based on the time-frequency transformation result after the zero-setting process, continue to extract the group delay ridge line of the mode with the maximum time-frequency energy to obtain the group delay ridge line of the second mode;

[0034] Perform group delay band filtering on the group delay ridge line of the second mode to remove noise interference and obtain the optimized group delay ridge line;

[0035] Set the time-frequency coefficients corresponding to the group delay ridge line of the optimized second mode to zero in the time-frequency transformation result;

[0036] Based on the time-frequency transformation result after the zero-setting process, continue to extract the group delay ridge line of the mode with the maximum time-frequency energy, and so on until the group delay ridge line of each mode is obtained.

[0037] In a feasible implementation manner, according to the group delay ridge line of each mode, obtain the optimized time-frequency transformation result, and perform time-frequency compression on the optimized time-frequency transformation result to obtain the optimized time-frequency image corresponding to the vibration signal, which specifically includes:

[0038] Substitute the group delay corresponding to the group delay ridge line of each mode into the time-frequency transformation result to obtain the optimized time-frequency transformation result;

[0039] Through the group delay compression transformation algorithm, perform transient compression transformation on the optimized time-frequency transformation result to obtain the optimized time-frequency image corresponding to the vibration signal.

[0040] In a feasible implementation manner, based on the group delay ridge line of each mode, diagnose the transient fault type of the current machine, which specifically includes:

[0041] Extract the impact interval frequency of the group delay ridge line;

[0042] Compare the impact interval frequency with the theoretical fault frequencies of various transient fault types to determine the transient fault type of the current machine; or,

[0043] Input the impact interval frequency and the optimized time-frequency image into a classifier for fault type diagnosis to obtain the transient fault type of the current machine.

[0044] On the other hand, the embodiment of the present invention also provides a mechanical transient fault diagnosis device based on group delay ridge line optimization, and the device includes:

[0045] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to execute the mechanical transient fault diagnosis method based on group delay ridge optimization.

[0046] Compared with the prior art, the mechanical transient fault diagnosis method and equipment based on group delay ridge optimization provided by the embodiments of the present invention have the following beneficial effects:

[0047] The present invention proposes to strengthen the stability of group delay estimation in the time-frequency processing algorithm based on the group delay ridge optimization technology, and utilize the central group delay and group delay bandwidth to realize the dynamic adjustment of the group delay ridge under strong frequency-varying signals and the elimination of noise in the time-frequency plane. At the same time, this process still maintains the ability to recover the original signal, realizes the improvement of the energy aggregation of the time-frequency representation and the increase of the anti-noise performance, so as to more accurately represent the group delay ridge impact characteristics during equipment failures and improve the accuracy of mechanical transient fault diagnosis. Description of the Drawings

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. In the drawings:

[0049] Figure 1 is a flowchart of a mechanical transient fault diagnosis method based on group delay ridge optimization provided by an embodiment of the present invention;

[0050] Figure 2 is a schematic diagram of a fitting function of a distribution histogram provided by an embodiment of the present invention;

[0051] Figure 3 is a schematic diagram of the structure of a mechanical transient fault diagnosis device based on group delay ridge optimization provided by an embodiment of the present invention. Detailed Embodiments

[0052] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0053] An embodiment of the present invention provides a mechanical transient fault diagnosis method based on group delay ridge optimization, as follows Figure 1 As shown, the mechanical transient fault diagnosis method based on group delay ridge optimization specifically includes steps S101 - S105:

[0054] S101. Perform time - frequency transformation and phase - modulated wavelet transformation on the collected vibration signal to obtain the corresponding time - frequency transformation result.

[0055] Specifically, model the collected vibration signal as a superposition of several intrinsic mode functions and perform Fourier transform to obtain a time - frequency signal model.

[0056] As a feasible implementation, model the collected vibration signal as a superposition of K frequency - varying intrinsic mode functions and perform Fourier transform. The obtained signal model is: ; where and respectively represent the instantaneous amplitude and instantaneous phase of the k - th mode in the signal, and the two can effectively characterize the energy and delay attributes of the signal varying with frequency. represents the frequency point, represents the frequency variable, represents the group delay of the signal, is the time - domain signal s ( t )'s Fourier transform.

[0057] Furthermore, perform phase modulation on the wavelet basis function through a scale factor and a translation factor, and perform wavelet transform on the time - frequency signal model through the modulated wavelet basis function to obtain the corresponding time - frequency transformation result.

[0058] As a feasible implementation, based on the above signal model, perform phase - modulated wavelet transform through a series of wavelet basis functions w() modulated by a scale factor and a translation factor to obtain the time - frequency transformation result: .

[0059] S102. Based on the time - frequency transformation result, extract the group delay ridge of the first mode in the vibration signal.

[0060] Specifically, divide the time - frequency plane where the time - frequency transformation result is located into several time - frequency segments along the frequency direction. Then, determine the point with the maximum time - frequency energy in each time - frequency segment as the search starting point, and search for the time - frequency energy peak points in the neighborhood to the left and right in turn. Finally, connect the point with the maximum time - frequency energy in the time - frequency segment and the time - frequency energy peak points to obtain the group delay ridge of each time - frequency segment.

[0061] Further, obtain the time-frequency energy of each time-frequency segment, and determine the time-frequency segment with the maximum time-frequency energy as the first mode; determine the group delay ridge line corresponding to the time-frequency segment with the maximum time-frequency energy as the group delay ridge line of the first mode.

[0062] S103. Perform group delay band filtering on the group delay ridge line to remove noise interference and obtain an optimized group delay ridge line.

[0063] Specifically, due to the influence of noise on the main components of the signal, the group delay estimation result obtained based on the time-frequency coefficients will have severe jitter near the ridge line. Therefore, the present invention optimizes and adjusts the group delay ridge line obtained based on the time-frequency coefficients, and the specific method is as follows:

[0064] First, take the time-frequency energy peak of the time-frequency transformation result as the group delay time-width center, and set the group delay ridge line at the group delay time-width center as the alignment reference value. The group delay time-width center can be expressed as: ; where s.t. represents the constraint condition.

[0065] Further, based on the alignment reference value, re-adjust the definition of the effective time-width region in the group delay estimation process as: ; where and are the left and right boundaries of the effective time-width region respectively, R represents the right boundary, L represents the left boundary, and C represents the group delay time-width center.

[0066] Further, based on the definition of the effective time-width region, determine the group delay estimator: .

[0067] Then, calculate the group delay band width of each group delay ridge line in the group delay estimator, count the number of each group delay band width, and draw a distribution histogram of the group delay band width.

[0068] Further, analyze the distribution histogram to obtain the filtering threshold: Fit the distribution histogram through the Trust-Region algorithm and analyze the distribution properties of the noise signal and the normal signal. Based on the distribution properties, obtain the intersection point of the noise signal distribution region and the normal signal distribution region, and use the group delay band width corresponding to the intersection point as the filtering threshold.

[0069] As a feasible implementation manner, Figure 2 is a schematic diagram of a fitting function of a distribution histogram provided by an embodiment of the present invention. As shown in Figure 2 is the fitting function corresponding to the distribution histogram fitted through the Trust-Region algorithm. The blue curve on the left is the noise curve, and the red curve on the right is the normal signal curve. The intersection point of the two curves is S t . Then use the time width S tAs the filtering threshold.

[0070] Furthermore, based on the filtering threshold, perform filtering on the group delay bands in the group delay ridge line to remove the noise interference in the group delay ridge line, and obtain the optimized group delay ridge line, specifically including:

[0071] Filter out the group delay bands in the group delay ridge line with a group delay band width less than the filtering threshold, and retain the group delay bands with a width greater than or equal to the filtering threshold, to obtain the filtered group delay ridge line.

[0072] S104. After removing the time-frequency coefficients corresponding to the optimized group delay ridge line, iteratively extract the group delay ridge line of the second mode, and perform ridge line optimization on the searched group delay ridge line; and so on, until the group delay ridge lines of each mode are obtained.

[0073] Specifically, set the time-frequency coefficients corresponding to the optimized group delay ridge line of the first mode to zero in the time-frequency transformation result.

[0074] Then, based on the time-frequency transformation result after the zero setting process, continue to extract the group delay ridge line of the mode with the maximum time-frequency energy to obtain the group delay ridge line of the second mode.

[0075] Continue to perform group delay band filtering on the group delay ridge line of the second mode to remove noise interference and obtain the optimized group delay ridge line. Set the time-frequency coefficients corresponding to the optimized group delay ridge line of the second mode to zero in the time-frequency transformation result.

[0076] Based on the time-frequency transformation result after the zero setting process, continue to extract the group delay ridge line of the mode with the maximum time-frequency energy, and so on, and cyclically extract the group delay ridge lines of each mode.

[0077] As a feasible implementation manner, in the process of adaptive detection of the group delay ridge line in each time-frequency segment, the key problem is how to set appropriate termination conditions in the ridge line detection process. For complex signals with significant differences in modal amplitudes and strong noise interference, introduce the group delay extraction operator: . This operator, as a binary operator, ignores the influence of the amplitude of the time-frequency coefficients and only has values at . Since GDO is a binary result consisting only of 0 and 1, and the group delay is distributed along the frequency direction as a single-valued function of frequency, the termination condition can be defined as: ; where is a constant threshold, and according to experience, it is recommended to set it to 0.9, and N is the length of the collected vibration signal.

[0078] S105, according to the group delay ridge of each mode, obtain the optimized time-frequency transformation result, and perform time-frequency compression on the optimized time-frequency transformation result to obtain the optimized time-frequency image corresponding to the vibration signal. Based on the group delay ridge of each mode, diagnose the transient fault type of the current machine.

[0079] Specifically, the group delay corresponding to the group delay ridge of each mode is substituted into the time-frequency transformation result to obtain the optimized time-frequency transformation result. Then, the group delay compression transformation function is used: , perform transient compression transformation on the optimized time-frequency transformation result to obtain the optimized time-frequency image corresponding to the vibration signal; where u represents time.

[0080] Furthermore, the impulse interval frequency of the group delay ridge is extracted. Then the impulse interval frequency is compared with the theoretical fault frequencies of various transient fault types to determine the transient fault type of the current machine; or, the impulse interval frequency and the optimized time-frequency image are input into the classifier for fault type diagnosis to obtain the transient fault type of the current machine.

[0081] As a feasible implementation method, the obtained optimized time-frequency processed image or the group delay slice image that needs to be identified can also be used as the input of the deep learning model to identify the fault problem of the equipment.

[0082] In addition, an embodiment of the present invention further provides a mechanical transient fault diagnosis device based on group delay ridge optimization, such as Figure 3 As shown, the mechanical transient fault diagnosis device based on group delay ridge optimization specifically includes:

[0083] at least one processor; and a memory in communication with the at least one processor; wherein,

[0084] The memory stores instructions executable by at least one processor to enable the at least one processor to perform:

[0085] Perform time-frequency transformation and phase modulation wavelet transformation on the collected vibration signal to obtain the corresponding time-frequency transformation result;

[0086] Based on the time-frequency transformation result, extracting the group delay ridge of the first mode in the vibration signal;

[0087] Performing group delay band filtering on the group delay ridge to remove noise interference and obtain an optimized group delay ridge;

[0088] After removing the time-frequency coefficients corresponding to the optimized group delay ridge, the group delay ridge of the second mode is extracted iteratively, and the searched group delay ridge is optimized; and so on, until the group delay ridge of each mode is obtained;

[0089] An optimized time-frequency transformation result is obtained according to the group delay ridge line of each mode, and time-frequency compression is performed on the optimized time-frequency transformation result to obtain an optimized time-frequency image corresponding to the vibration signal;

[0090] Based on the group delay ridge line of each mode, diagnose the transient fault type of the current machine.

[0091] Each embodiment in the present invention is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments. In particular, for the embodiments of the device, equipment, and non-volatile computer storage medium, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiments.

[0092] The above describes specific embodiments of the present invention. Additionally, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0093] The above are only the embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, various changes and modifications can be made to the embodiments of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of the present invention shall be included within the protection scope of the present invention.

Claims

1. A mechanical transient fault diagnosis method based on group delay ridge optimization, characterized in that: The method comprises: Perform time-frequency transformation and phase modulation wavelet transformation on the collected vibration signal to obtain the corresponding time-frequency transformation result; Based on the time-frequency transformation result, extracting the group delay ridge of the first mode in the vibration signal specifically includes: The time-frequency plane where the time-frequency transform result is located is divided into a number of time-frequency segments along the frequency direction; the maximum time-frequency energy point in each time-frequency segment is determined as the search starting point, and the time-frequency energy peak points in the neighborhood are searched to the left and right in turn; the maximum time-frequency energy point in the time-frequency segment is connected with the time-frequency energy peak point to obtain the group delay ridge of each time-frequency segment; the time-frequency energy of each time-frequency segment is obtained, and the time-frequency segment with the maximum time-frequency energy is determined as the first mode; the group delay ridge corresponding to the time-frequency segment with the maximum time-frequency energy is determined as the group delay ridge of the first mode; The group delay ridge is subjected to group delay band filtering to remove noise interference and obtain an optimized group delay ridge, specifically comprising: The time-frequency energy peak of the time-frequency transformation result is used as the center of the group delay time width, and the group delay ridge at the center of the group delay time width is set as the alignment reference value; based on the alignment reference value, the effective time width area definition in the group delay estimation process is readjusted; based on the effective time width area definition, a group delay estimator is determined; each group delay band width of the group delay ridge in the group delay estimator is calculated, and the number of each group delay band width is counted, and a distribution histogram of the group delay band width is drawn; the distribution histogram is analyzed to obtain a filtering threshold; based on the filtering threshold, the group delay band in the group delay ridge is filtered to remove noise interference in the group delay ridge to obtain an optimized group delay ridge; After removing the time-frequency coefficients corresponding to the optimized group delay ridge, the group delay ridge of the second mode is iteratively extracted, and the searched group delay ridge is optimized; and so on, until the group delay ridge of each mode is obtained; According to the group delay ridge of each mode, an optimized time-frequency transformation result is obtained, and the optimized time-frequency transformation result is time-frequency compressed to obtain an optimized time-frequency image corresponding to the vibration signal; Based on the group delay ridge of each mode, the transient fault type of the current machinery is diagnosed.

2. A mechanical transient fault diagnosis method based on group delay ridge optimization according to claim 1, characterized in that: The collected vibration signal is subjected to time-frequency transformation and phase modulation wavelet transformation to obtain the corresponding time-frequency transformation results, including: The collected vibration signal is modeled as a superposition of several intrinsic mode functions and Fourier transformed to obtain a time-frequency signal model; The wavelet basis function is phase modulated by the scale factor and the translation factor, and the time-frequency signal model is subjected to wavelet transform by the modulated wavelet basis function to obtain a corresponding time-frequency transform result.

3. The mechanical transient fault diagnosis method based on group delay ridge optimization according to claim 1 is characterized in that: Analyzing the distribution histogram to obtain a filtering threshold specifically includes: Fitting the distribution histogram by the Trust-Region algorithm to analyze the distribution properties of the noise signal and the normal signal; Based on the distribution property, the intersection point of the noise signal distribution area and the normal signal distribution area is obtained, and the group delay bandwidth corresponding to the intersection point is used as the filtering threshold.

4. The mechanical transient fault diagnosis method based on group delay ridge optimization according to claim 1 is characterized in that: Based on the filtering threshold, filtering the group delay band in the group delay ridge includes: Among the group delay ridges, group delay bands whose width is less than the filtering threshold are filtered out, and group delay bands whose width is greater than or equal to the filtering threshold are retained to obtain the filtered group delay ridges.

5. The mechanical transient fault diagnosis method based on group delay ridge optimization according to claim 1 is characterized in that: After removing the time-frequency coefficients corresponding to the optimized group delay ridge, the group delay ridge of the second mode is iteratively extracted, and the searched group delay ridge is optimized; And so on, until the group delay ridge of each mode is obtained, including: Setting the time-frequency coefficient corresponding to the optimized group delay ridge of the first mode to zero in the time-frequency transform result; Based on the time-frequency transformation result after zeroing, the group delay ridge of the mode with the largest time-frequency energy is further extracted to obtain the group delay ridge of the second mode; Performing group delay band filtering on the group delay ridge of the second mode to remove noise interference and obtain an optimized group delay ridge; Setting the time-frequency coefficient corresponding to the optimized group delay ridge of the second mode to zero in the time-frequency transform result; Based on the time-frequency transformation result after the zeroing process, the group delay ridge of the mode with the largest time-frequency energy continues to be extracted, and so on, until the group delay ridge of each mode is obtained.

6. The mechanical transient fault diagnosis method based on group delay ridge optimization according to claim 1, characterized in that: According to the group delay ridge of each mode, an optimized time-frequency transformation result is obtained, and the optimized time-frequency transformation result is time-frequency compressed to obtain an optimized time-frequency image corresponding to the vibration signal, specifically including: Substituting the group delay corresponding to the group delay ridge of each mode into the time-frequency transformation result to obtain an optimized time-frequency transformation result; The optimized time-frequency transformation result is subjected to transient compression transformation by means of a group delay compression transformation algorithm to obtain an optimized time-frequency image corresponding to the vibration signal.

7. The mechanical transient fault diagnosis method based on group delay ridge optimization according to claim 1 is characterized in that: Based on the group delay ridge of each mode, the transient fault type of the current machinery is diagnosed, including: extracting the impulse interval frequency of the group delay ridge; Compare the impact interval frequency with the theoretical fault frequencies of various transient fault types to determine the transient fault type of the current machine; or, The impact interval frequency and the optimized time-frequency image are input into a classifier to perform fault type diagnosis, thereby obtaining the transient fault type of the current machine.

8. A mechanical transient fault diagnosis device based on group delay ridge optimization, characterized in that: The device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, so that the at least one processor can execute the mechanical transient fault diagnosis method based on group delay ridge optimization according to any one of claims 1-7.

Images