Machine pump bearing weak fault diagnosis method and system based on time-frequency adaptive segmentation

Abstract

The invention provides a machine pump bearing weak fault diagnosis method and system based on time-frequency self-adaptive segmentation. The method comprises the steps that collected vibration signals are converted and processed to obtain corresponding time-frequency distribution characteristics; quantitatively representing the adjacent correlation distribution relation of the time-frequency distribution characteristics, and realizing frequency domain interval division to obtain multiple parts of time-frequency distribution characteristics; determining fault characteristic frequency ratio indexes corresponding to time-frequency distribution characteristics of different parts by using a principal component analysis technology, and selecting sensitive components; and performing coefficient optimization processing by adopting a differential search algorithm, calculating to obtain a fault feature component, and extracting and diagnosing weak fault features of the pump bearing based on the fault feature component. By adopting the scheme, the problems of single theory, complex calculation and insufficient accuracy in the prior art can be solved, the weak fault diagnosis of equipment is realized by comprehensively utilizing the time-frequency adaptive segmentation technology and the frequency band coefficient optimization algorithm, the timeliness is good, the operation result is finer, and the recognition and diagnosis of the weak fault of the bearing are timely and reliably realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of predictive maintenance of mechanical equipment, and particularly relates to a weak fault diagnosis method and system for bearings of a machine pump based on time-frequency adaptive segmentation. BACKGROUND

[0002] The bearing is an important component in the machine pump, and the alternating load caused by the severe working environment and time-varying working condition is easy to cause the bearing to fail, and if no measures are taken in time, a serious accident will be caused and great loss will be caused. When the bearing has local damage such as fatigue spalling and crack, the defect part will excite a periodic one-sided decay vibration signal in the process of contacting with other components of the bearing, the periodic frequency is the characteristic frequency of different element faults, and the one-sided decay signal vibration frequency is the inherent frequency of the bearing-sensor system.

[0003] Because there is strong noise interference in the low frequency band, it is difficult to capture the bearing fault information by directly analyzing the frequency spectrum of the vibration signal, therefore, the bearing fault diagnosis generally starts from the high frequency part of the vibration signal, selects the resonance frequency band with obvious bearing fault characteristics for analysis, separates the inherent vibration through the band-pass filter with the center frequency equal to the inherent frequency, removes the frequency components of the high-frequency decay vibration through the Hilbert transform for envelope demodulation, obtains the low-frequency envelope signal containing only the fault characteristic information, and analyzes the frequency domain distribution of the envelope signal to judge the fault type of the bearing. The most important step in the envelope demodulation method is to segment the signal in the frequency domain and select an appropriate frequency band for analysis, but the signal processing method used in the prior art generally starts from the frequency domain angle; such signal processing method has problems of complex calculation, too many preset parameters, insufficient accuracy of processing result, and is easy to be disturbed by noise or non-related signal components.

[0004] The information disclosed in the background section of this application is only intended to deepen the understanding of the general background of the application, and should not be regarded as recognizing or implying in any form that this information constitutes the prior art known to those skilled in the art. SUMMARY

[0005] To solve the above problems, the application provides a weak fault diagnosis method for machine pump bearing based on time-frequency adaptive segmentation, which transforms and processes the collected vibration signal to obtain corresponding time-frequency distribution characteristics; quantizes the adjacent correlation distribution relationship of the time-frequency distribution characteristics to realize frequency domain interval division and obtain time-frequency distribution characteristics in multiple parts; uses principal component analysis technology to determine the fault characteristic frequency ratio index corresponding to the time-frequency distribution characteristics in different parts, and selects sensitive components; further uses a differential search algorithm for coefficient optimization processing to calculate and obtain fault characteristic components, and realizes weak fault feature extraction and diagnosis of machine pump bearing based on the same. This scheme can overcome the problems of single theory, complex calculation and insufficient accuracy in the prior art, and comprehensively utilizes time-frequency adaptive segmentation technology and frequency band coefficient optimization algorithm to realize weak fault diagnosis of equipment, which is good in timeliness and more accurate in calculation results. Preferably, in one embodiment, the method comprises:

[0006] a signal processing step of transforming and processing the collected vibration signal to obtain corresponding time-frequency distribution characteristics;

[0007] a feature division step of quantizing the adjacent correlation distribution relationship of the time-frequency distribution characteristics to realize frequency domain interval division and obtain time-frequency distribution characteristics in multiple parts;

[0008] a sensitive frequency band determination step of using principal component analysis technology to determine the fault characteristic frequency ratio index corresponding to the time-frequency distribution characteristics in different parts, and selecting sensitive components;

[0009] an adaptive optimization processing step of using a differential search algorithm for optimization processing based on the fault characteristic frequency ratio index to obtain fault characteristic components, and realizing weak fault feature extraction and diagnosis of machine pump bearing based on the same.

[0010] Optionally, in one embodiment, in the signal processing step, the time-frequency distribution characteristics are calculated based on the collected machine pump bearing vibration signal using short-time Fourier transform.

[0011] Further, in one embodiment, in the signal processing step, the time-frequency distribution characteristics of the vibration signal are calculated according to the following formula:

[0012]

[0013] In the formula, STFT(t, f) represents the time-frequency distribution characteristics of the signal, f represents the frequency axis, s(τ) represents the vibration signal of the machine pump equipment, h(τ-t) represents the window function for segmenting the input signal for short-time Fourier transform, and τ represents the time parameter inside the window function.

[0014] In a preferred embodiment, in the signal processing step, in the process of calculating the corresponding time-frequency distribution characteristics, a window function with a fixed time-frequency shape is used, the resolution in the time-frequency domain does not change with time and frequency, and the corresponding time-frequency grid size is fixed.

[0015] In one embodiment, in the feature division step, the distribution similarity of adjacent components in the frequency dimension is calculated to quantify the adjacent correlation distribution relationship that characterizes the time-frequency distribution features.

[0016] Furthermore, in one embodiment, the distribution similarity of adjacent components in the frequency dimension is calculated as follows:

[0017]

[0018] Among them, STFT(τ,f) and STFT(τ,f+1) represent the time-frequency distribution of two adjacent frequency axes, frequency axis f and frequency axis f+1, respectively, and Corr(f) represents the similarity of adjacent correlation distribution of time-frequency distribution features.

[0019] Optionally, in one embodiment, in the feature division step, the midpoint between two adjacent maximum points of adjacent correlation distribution similarities is used as the division boundary to adaptively decompose the frequency domain support interval of the signal to be processed into multiple parts to form a multi-part time-frequency distribution feature.

[0020] In a preferred embodiment, the sensitive frequency band determination step includes the following operations:

[0021] The principal component analysis technique is used to obtain the principal components of each time-frequency part and obtain the spectrum of the principal components of each time-frequency part;

[0022] The fault frequency ratio index is used to measure the fault feature information contained in each time-frequency part, and the time-frequency component with the largest fault feature frequency ratio index is selected as the fault sensitive component.

[0023] Furthermore, in one embodiment, the principal component of each time-frequency part is obtained according to the following logic:

[0024] P(t,n)=PCA[STFT(t,Λ n )]n=1,2,…,N

[0025] Among them, P(t,n) represents the principal component corresponding to the nth part of the time-frequency signal, STFT(t,Λ n ) represents the time-frequency signal of the nth part, and N represents the time-frequency signal of N parts in total.

[0026] Optionally, in one embodiment, in the adaptive optimization processing step, the coefficients of each frequency axis of the sensitive part are adaptively optimized using the obtained fault characteristic frequency ratio index as a fitness function.

[0027] Further, in one embodiment, the process of adaptive optimization includes:

[0028] determining the number of population individuals and the maximum number of optimization iterations;

[0029] initializing the position of the individual, and calculating the fitness of the current position of the individual according to the fitness function;

[0030] calculating the temporary stop position of all individuals, and if it exceeds the boundary of the optimization space, moving it to a random position within the optimization space, and calculating the corresponding fitness;

[0031] comparing the fitness of the temporary stop position of the individual with the fitness of the current position, and if the former is better, moving the position of the individual to the temporary stop position and updating the fitness, otherwise keeping the original position;

[0032] after all individuals perform the deletion operation, updating the population and the global optimal position; and performing a loop operation until a set requirement is met to output the global optimal position and its fitness, and obtaining the corresponding optimization coefficient as the target coefficient.

[0033] Based on other aspects of the method described in any one or more of the above embodiments, the application also provides a storage medium having program codes stored thereon, which can implement the method described in any one or more of the above embodiments.

[0034] Based on the application aspects of the method described in any one or more of the above embodiments, the application also provides a weak fault diagnosis system for machine pump bearing based on time-frequency adaptive segmentation, which performs the method described in any one or more of the above embodiments.

[0035] Compared with the closest prior art, the application also has the following beneficial effects:

[0036] The application provides a weak fault diagnosis method and system for machine pump bearing based on time-frequency adaptive segmentation, which transforms and processes the collected vibration signal to obtain corresponding time-frequency distribution characteristics; quantifies the adjacent correlation distribution relationship representing the time-frequency distribution characteristics, realizes frequency domain interval division to obtain multiple time-frequency distribution characteristics, and comprehensively analyzes the vibration signal in the joint distribution of time domain-frequency domain, so that the analysis of the vibration signal is more comprehensive, the relationship between each frequency band is judged by using the distribution manifold of the vibration signal in the time-frequency domain, adaptive decomposition of different frequency band signals is realized, and the decomposition result is more reasonable.

[0037] Further, the principal component analysis technology is used to determine the fault characteristic frequency ratio indexes corresponding to different part time-frequency distribution characteristics, and sensitive components are selected; further, the differential search algorithm is used for coefficient optimization processing, and the fault characteristic components are calculated and obtained, and the weak fault characteristic extraction and diagnosis of the machine pump bearing are realized based on the fault characteristic components; the scheme selects the fault sensitive frequency band by using the fault characteristic frequency ratio indexes of the principal component spectrums of the decomposed parts, and on the basis, the adaptive optimization of the frequency axis coefficients of the sensitive frequency band is combined with the differential search algorithm, the bearing fault characteristics are further enhanced, the influence of other components in the gearbox is eliminated, the accurate weak fault diagnosis is realized, the operation process is simple, and the timeliness is good; the scheme provides guidance for predictive maintenance of the dynamic equipment, and guarantees long-term stable operation and safe industrial production of the device.

[0038] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be realized and attained by the structure particularly pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0039] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, illustrate embodiments of the present application, and are used to explain the present application together with the description. The present application is not limited by the accompanying drawings.

[0040] Figure 1 is a flowchart of a machine pump bearing weak fault diagnosis method based on time-frequency adaptive segmentation provided by an embodiment of the present application;

[0041] Figure 2 is a time domain distribution diagram of a vibration signal in the machine pump bearing weak fault diagnosis method based on time-frequency adaptive segmentation provided by the embodiment of the present application;

[0042] Figure 3 is a time-frequency domain diagram of a vibration signal in the machine pump bearing weak fault diagnosis method based on time-frequency adaptive segmentation provided by the embodiment of the present application;

[0043] Figure 4 is a time-frequency adjacent correlation distribution diagram of a vibration signal in the machine pump bearing weak fault diagnosis method based on time-frequency adaptive segmentation provided by the embodiment of the present application;

[0044] Figure 5 is a time-frequency component distribution diagram of a vibration signal after segmentation in the machine pump bearing weak fault diagnosis method based on time-frequency adaptive segmentation provided by the embodiment of the present application;

[0045] Figure 6 is a sensitive component principal component spectrum diagram example in the machine pump bearing weak fault diagnosis method based on time-frequency adaptive segmentation provided by the embodiment of the present application;

[0046] Figure 7 is an example of a combined component spectrum corresponding to an adaptive optimization coefficient in a method for diagnosing weak faults of a pump bearing based on time-frequency adaptive segmentation provided by an embodiment of the present application. DETAILED DESCRIPTION

[0047] The embodiments of the present application will be described in detail hereinafter with reference to the drawings and embodiments, by which the person skilled in the art can fully understand how the present application applies technical means to solve technical problems and achieve the implementation process of technical effects and implement the present application according to the above implementation process. It should be noted that, as long as there is no conflict, each embodiment in the present application and each feature of each embodiment can be combined with each other, and the technical solutions formed thereby are all within the protection scope of the present application.

[0048] Although the flowcharts describe the operations as sequential processes, many of the operations can be performed in parallel, concurrently or at the same time. The order of the operations can be re-arranged. The process can be terminated when its operations are completed, but can also have additional steps not included in the figure. The process can correspond to methods, functions, routines, subroutines, subprograms, etc.

[0049] The computer device includes a user device and a network device. The user device or client includes, but is not limited to, a computer, a smart phone, a PDA (Personal Digital Assistant), etc.; the network device includes, but is not limited to, a single network server, a server group composed of multiple network servers, or a cloud composed of a large number of computers or network servers based on cloud computing. The computer device can be independently operated to implement the present application, or can be connected to a network and interact with other computer devices in the network to implement the present application. The network in which the computer device is located includes, but is not limited to, the Internet, a wide area network, a metropolitan area network, a local area network, a VPN network, etc.

[0050] The terms "first", "second", and the like can be used to describe various elements, but these elements should not be limited by these terms. These terms are used only to distinguish one element from another. The term "and / or" as used herein includes any and all combinations of one or more of the associated associated items. When a unit is referred to as "connected" or "coupled" to another unit, it can be directly connected or coupled to the other unit, or there can be an intermediate unit.

[0051] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0052] Bearing is an important component in the machine pump, the alternating load caused by the bad working environment and time-varying working condition is easy to cause the bearing failure, if not take measures in time will cause serious accident and cause great loss.

[0053] When the bearing occurs fatigue spalling, crack and other local damage, the defect site will excite periodic one-sided decay vibration signal in the process of contacting with other components of the bearing, the periodic frequency is the characteristic frequency of different element failure, and the one-sided decay signal vibration frequency is the inherent frequency of bearing-sensor system.

[0054] Because of the strong noise interference in the low frequency band, it is difficult to capture the bearing fault information by directly analyzing the frequency spectrum of the vibration signal, therefore, the bearing fault diagnosis generally starts from the high frequency part of the vibration signal, selects the resonance frequency band with obvious bearing fault characteristics for analysis, separates the inherent vibration through the band-pass filter with the center frequency equal to the inherent frequency, removes the frequency components of the high frequency decay vibration by using Hilbert transform for envelope demodulation, obtains the low frequency envelope signal containing only the fault characteristic information, and analyzes the frequency domain distribution of the envelope signal to determine the bearing fault type.

[0055] The most important step in the envelope demodulation method is to divide the signal in the frequency domain and select the appropriate frequency band for analysis, but the signal processing method used in the prior art generally starts from the frequency domain alone. For example, patent document CN202210295430.5 proposes a bearing weak fault diagnosis method, which decomposes the vibration signal using the continuous variational mode decomposition method to obtain multiple modes containing single signal components; the modes are sorted according to the size of the sparsity calculated according to the formula; according to the sorting result, the signal is reconstructed based on the sparsity index of the reorganization strategy; the reconstructed signal is analyzed to extract the weak fault features of the bearing. Patent document CN202210277235.X provides a rolling bearing weak fault diagnosis method based on PSO-VMD-MCKD, which optimizes the VMD algorithm using the PSO algorithm, selects the optimal modal component based on the VMD decomposition result of the weak fault signal, and determines the optimization range of T in the MCKD algorithm based on the prominent frequency range in the envelope spectrum of the optimal modal component. The MCKD algorithm is optimized using PSO, and the fault impact component in the optimal component signal is strengthened based on the MCKD algorithm. Finally, the envelope spectrum extracts the weak fault features of the bearing and compares them with the theoretical fault frequency to obtain the fault diagnosis result. CN202110631601.2 provides a high-speed train wheel bearing weak fault diagnosis method and system, which uses the signal feature frequency update trend of the high-speed train wheel bearing to construct a frequency spectrum feature information scanner to obtain the boundary parameters of the potential feature components in the signal; a filter bank is constructed according to the boundary parameters to decompose the high-speed train wheel bearing signal and obtain the corresponding mode components; a fusion fault sensitive index is established to locate the fault feature components by calculating the fault sensitive index of the mode components; envelope analysis is performed on the fault feature components to obtain an envelope spectrum, and the weak fault features are detected according to the envelope spectrum to complete the weak fault diagnosis of the high-speed train wheel bearing.

[0056] The signal processing method of the above prior art has problems such as complex calculation, too many preset parameters, insufficient accuracy of processing results, and is easily disturbed by noise or non-related signal components.

[0057] To solve the above problems, the present application provides a bearing weak fault diagnosis method and system based on time-frequency adaptive segmentation, which is based on a vibration signal decomposition method related to the adjacent time-frequency distribution, uses the distribution of the vibration signal in the time-frequency domain to judge the relationship between the frequency bands, and then realizes the adaptive decomposition of the signals in different frequency bands; on this basis, the fault sensitive frequency band is selected using the fault feature frequency ratio index of the principal component spectrum obtained by decomposition, and the differential search algorithm is used to adaptively optimize the frequency axis coefficients of the sensitive frequency band, thereby realizing the diagnosis of the bearing weak fault

[0058] Next, the detailed flow of the method of the embodiment of the application is described in detail based on the drawings, the steps shown in the flowchart of the drawings can be executed in a computer system comprising computer executable instructions such as a set of computer executable instructions. Although the logical order of the steps is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown.

[0059] Embodiment one:

[0060] The researchers of the application consider that envelope analysis is a standard technique for fault diagnosis, which needs to select a specific frequency band by designing a band-pass filter and the like, and then calculate the envelope of the signal in the selected frequency band by using a Hilbert transform or the like to realize positioning and identification of faults. Based on this, how to segment the frequency domain of the vibration signal in high quality and select a suitable frequency band for demodulation is the key to accurate identification of faults by using envelope analysis.

[0061] However, the existing signal processing method generally starts from the frequency domain, and has problems such as complex calculation, too many preset parameters, and is easily disturbed by noise or non-related signal components. It is difficult to accurately determine the signal frequency domain segmentation boundary, which has an adverse effect on the analysis and diagnosis of bearing weak faults.

[0062] The application provides a device weak fault diagnosis technology based on vibration signal time-frequency adaptive segmentation and frequency band coefficient optimization. Firstly, the mutual relationship between each frequency band of the signal is obtained according to the adjacent correlation information of the signal time-frequency distribution, and the signal time-frequency distribution is segmented; then, the fault sensitive frequency band is selected by using the fault characteristic frequency ratio of the principal component spectrum of each part obtained by decomposition as an index, and on this basis, the adaptive optimization of each frequency axis coefficient of the sensitive frequency band is realized by using a differential search algorithm, so that the diagnosis of bearing weak faults is realized.

[0063] Figure 1 A flowchart of a bearing weak fault diagnosis method based on time-frequency adaptive segmentation provided by the embodiment one of the application is shown, and the method comprises the following steps. Figure 1 It can be seen that in the preferred embodiment, the method comprises the following steps.

[0064] The signal processing step: the collected vibration signal is transformed and processed to obtain corresponding time-frequency distribution characteristics;

[0065] The characteristic division step: the adjacent correlation distribution relationship representing the time-frequency distribution characteristics is quantified, the frequency domain interval is divided, and the time-frequency distribution characteristics in a multi-part form are obtained;

[0066] The sensitive frequency band determination step: the fault characteristic frequency ratio index corresponding to the different part time-frequency distribution characteristics is determined by using the principal component analysis technology, and the sensitive component is selected;

[0067] Adaptive optimization processing step: based on the fault characteristic frequency ratio index, a differential search algorithm is used for optimization processing to obtain a fault characteristic component, and based on the fault characteristic component, bearing weak fault feature extraction and diagnosis of a machine pump shaft are realized.

[0068] The machine pump bearing weak fault diagnosis method based on time-frequency adaptive segmentation provided by the embodiment of the application first obtains the mutual relationship between each frequency band of a signal according to adjacent correlation information of signal time-frequency distribution, segments the signal time-frequency distribution, and selects a fault sensitive frequency band by using a fault characteristic frequency ratio index of a principal component spectrum obtained by decomposition.

[0069] The embodiment of the application first transforms and processes the collected vibration signal to obtain corresponding time-frequency distribution features by the signal processing step.

[0070] For example, in one embodiment, when the vibration signal is collected, a DT9837 data sampling card is used to collect the vibration signal at the support bearing, the sampling frequency is 20480 Hz, and the sampling time length is 1 s. Figure 2 According to the current information, no obvious periodic impact can be observed from the figure, the bearing fault characteristic information is covered by the vibration signal of other components and noise, and further processing and mining are needed.

[0071] The embodiment of the application obtains the time-frequency distribution processing corresponding to the machine pump bearing vibration signal by the signal processing step, the basic idea of time-frequency analysis processing is to convert the vibration signal to the time-frequency domain for analysis and processing, and the characteristics of the signal are observed in the time and frequency dimensions. The application uses short-time Fourier transform to calculate the signal time-frequency distribution. Further time-frequency transform, in order to extract the bearing fault characteristics, the vibration signal is mapped to the time-frequency domain for processing by short-time Fourier transform, in order to make the bearing periodic impact information more obvious, according to the foregoing sampling frequency and bearing structure, the application selects a window length of 300 points, and the vibration signal time-frequency distribution is as shown in Figure 3 At this time, the time-frequency distribution is still relatively complex and needs further processing and analysis

[0072] Therefore, optionally, in one embodiment, in the signal processing step, the short-time Fourier transform is used to calculate the corresponding time-frequency distribution features of the collected machine pump bearing vibration signal.

[0073] Further, in one optional embodiment, the time-frequency distribution features of the vibration signal are calculated according to the following formula:

[0074]

[0075] In the formula, STFT(t,f) represents the time-frequency distribution characteristics of the signal, f represents the frequency axis, s(τ) represents the vibration signal of the pump equipment, h(τ-t) represents a window function for segmenting the input signal for short-time Fourier transform, and τ represents a time parameter inside the window function.

[0076] The STFT segments the signal for FFT (fast Fourier transform) through the window function h(τ-t), and the parameter τ is used to shift the window function on the time axis to cover the entire time domain. The STFT uses a window function h(t) with a fixed time-frequency shape in the entire signal processing process, and the resolution in the time-frequency domain does not change with time and frequency, that is, a fixed time-frequency grid is used.

[0077] That is, in an optional embodiment, in the signal processing step, the process of calculating the corresponding time-frequency distribution characteristics uses a window function with a fixed time-frequency shape, the resolution in the time-frequency domain does not change with time and frequency, and the corresponding time-frequency grid size is fixed.

[0078] Because of different generation mechanisms and transmission paths, the vibration signals generated by different parts of the rotating machinery are often distributed in different time-frequency regions, and especially show the characteristics of aggregation in the frequency dimension. For example, the vibration information of gears and rotors is often concentrated in the low-frequency part of the signal, and the characteristics of bearing vibration signals are often more obvious in the high-frequency part.

[0079] Therefore, further, by the feature division step, the adjacent correlation distribution relationship representing the time-frequency distribution characteristics is quantified, the frequency domain boundary is divided, and the time-frequency distribution characteristics in the form of multiple parts are obtained.

[0080] In the feature division step, the distribution similarity of adjacent components in the frequency dimension is calculated to quantify the adjacent correlation distribution relationship representing the time-frequency distribution characteristics.

[0081] Specifically, in one embodiment, the distribution similarity of adjacent components in the frequency dimension is calculated according to the following formula to represent the adjacent correlation distribution in the frequency dimension of the time-frequency distribution:

[0082]

[0083] Wherein, STFT(τ,f) and STFT(τ,f+1) represent the time-frequency distributions of two adjacent frequency axes f and f+1, respectively, and Corr(f) represents the adjacent correlation distribution similarity of the time-frequency distribution characteristics. If the distribution similarity of adjacent frequency axes is higher, the value of Corr(f) is larger, and this can be used as the basis for time-frequency adaptive segmentation. The time-frequency adjacent correlation distribution of the vibration signal is shown in FIG. 1, from which six local maximum values can be observed. Figure 4

[0084] ​Based on this, in the feature division step, further using the midpoint between two adjacent maximum points of the adjacent correlation distribution as the division boundary, the frequency domain support interval of the to-be-processed signal is adaptively decomposed into multiple parts, forming a time-frequency distribution feature in the form of multiple parts.

[0085] In actual application, the frequency-correlation degree relationship diagram can be drawn according to the calculated distribution similarity of adjacent frequency axes, and then the frequency domain support interval [0, Fs / 2] of the to-be-processed signal is adaptively decomposed into N parts, and preferably, the frequency coordinate at the midpoint between two adjacent maximum points of the signal adjacent correlation distribution is used to represent the frequency domain boundary between each part, as shown in Figure 4 .

[0086] Λ n = [ω n-1 ,ω n ], n = 1, 2,..., N

[0087]

[0088] Wherein, Λ n represents the nth part of the divided signal time-frequency feature

[0089] Next, the sensitive frequency band determination step is performed: using principal component analysis technology to determine the fault feature frequency ratio index corresponding to the different part time-frequency distribution features, and selecting the sensitive component.

[0090] Optionally, in an embodiment, in the sensitive frequency band determination step, the following operations are included:

[0091] The principal component P of each time-frequency part is calculated using principal component analysis technology:

[0092] P(t, n) = PCA[STFT(t, Λ n )] n = 1, 2,..., N

[0093] In the formula, P(t, n) represents the principal component corresponding to the nth part of the time-frequency signal, STFT(t, Λ n ) represents the time-frequency signal of the nth part, and N represents the total of N time-frequency signals.

[0094] The essence of the principal component analysis algorithm in the above formula is to obtain the coefficients of each component α ω according to the mutual relationship of the high-dimensional components, and then obtain the principal component by weighted summation of each part; the higher the distribution similarity of adjacent frequency axes, the larger the value of α ω . Further, the spectrum of each time-frequency part principal component is obtained

[0095]

[0096] In the formula, represents the spectral amplitude of the nth part corresponding to the principal component, and n represents the nth principal component

[0097] A fault frequency ratio index is calculated, the fault feature information contained in each time-frequency part is measured by the fault frequency ratio index, and a time-frequency component with the maximum fault feature frequency ratio index is selected as a fault sensitive component.

[0098]

[0099] In the formula, R ω represents the fault frequency ratio index; represents the amplitude corresponding to the fault frequency, represents the total sum of the full spectrum amplitudes.

[0100] The fault frequency ratio index distribution of the analyzed signal is as shown in Figure 5 The sixth part with the highest fault frequency ratio index value of the principal component (principal component) spectrum is defined as a fault sensitive frequency band. The sixth part corresponding to the principal component spectrum is as shown in Figure 6 From which the bearing inner ring fault feature frequency 78 Hz can be observed.

[0101] The aforementioned principal component analysis only solves from the perspective of covariance when calculating the coefficients of each frequency axis, and does not fully mine the fault information in the signal; in order to further enhance the bearing fault feature and eliminate the influence of other components in the gearbox, a differential search algorithm is introduced to adaptively optimize the coefficients of each frequency axis of the sixth part with the fault frequency ratio index as the fitness function.

[0102] Therefore, an adaptive optimization processing step is performed, the differential search algorithm is used to optimize the processing based on the fault feature frequency ratio index, and the fault feature component is obtained, based on which the weak fault feature extraction and diagnosis of the bearing of the machine pump are realized.

[0103] In the adaptive optimization processing step, the obtained fault feature frequency ratio index is used as the fitness function to adaptively optimize the coefficients of each frequency axis of the sensitive part.

[0104] On this basis, the equipment fault frequency ratio index of the spectrum of the optimal fault feature component D(t, n) is used as the fitness function in the optimization process, and the differential search algorithm is used to optimize the coefficients β ω of each frequency axis of the sensitive component, and the optimal fault feature component D(t, n) is obtained:

[0105]

[0106] In an optional embodiment, the adaptive optimization process includes:

[0107] The number of population individuals N and the maximum optimization iteration number G are determined.

[0108] Initialize the position of individual, calculate the fitness of the current position of individual according to the fitness function;

[0109] Calculate the temporary stop position of all individuals, if it exceeds the boundary of the optimization space, move it to a random position within the optimization space, and calculate the corresponding fitness;

[0110] Compare the fitness of the temporary stop position of individual with the fitness of the current position, if the former is better, move the position of individual to the temporary stop position and update the fitness, otherwise keep the original position;

[0111] After all individuals perform the deletion operation, update the population and the global optimal position; loop until the set requirement is output, and obtain the corresponding optimization coefficient as the target coefficient.

[0112] After obtaining the coefficient, combine the frequency axis information of the sixth part by weighting, obtain the weighted component spectrum, and perform weak fault feature extraction and diagnosis according to the weighted component spectrum.

[0113] In actual application, the specific steps are as follows:

[0114] ① Determine the number of population individuals N and the maximum optimization iteration number G;

[0115] ② Initialize the position of individual by using the following formula, the optimization coefficient range is [0, 1], and calculate the fitness of the current position of individual according to the fitness function device fault frequency ratio index;

[0116] x i,j =rand(up j -low j )+low j

[0117] x i,j represents the jth element of the ith individual of the population; rand represents a random number between 0 and 1 satisfying uniform distribution; up j represents the upper limit of the jth dimension of the search frequency domain boundary space; low j represents the lower limit of the jth dimension of the search frequency domain boundary space.

[0118] ③ Calculate the temporary stop position of all individuals according to the following formula, if the temporary stop position exceeds the boundary [0, 1] of the optimization space, move it to a random position within the optimization space;

[0119] S=X i +γ·(donor-X i )

[0120] In the formula, X iIndicates the current location of the population, donor indicates the population movement target, and γ is a random number that satisfies the gamma distribution.

[0121] ④ Compare the fitness functions of the individual's temporary location and current location to determine whether the fitness of the temporary location is better than the current location. If the temporary location is better than the current location, the individual immediately moves to the temporary location and updates its fitness, otherwise it stays at the original location. After completion, the individual waits for other individuals in the population to complete this step. When all individuals have completed the exploration and comparison of the temporary location, the population and the global optimal position are updated;

[0122]

[0123] In the formula, fit represents the fitness function, S represents the temporary stay position, S * Indicates the current location.

[0124] ⑤If the specified total number of iterations is not reached, return to step ③. If the total number of iterations is reached, output the global optimal position and its fitness.

[0125] After obtaining the coefficients, the frequency axis information of the sixth part is weighted and combined, and the combined component spectrum is as follows: Figure 7 As shown, compared Figure 6 As shown in the principal component spectrum, the bearing fault characteristics are significantly enhanced.

[0126] The present invention is further described below with reference to examples of implementation. The scope of the present invention is not limited by the examples, but is set forth in the claims.

[0127] The specific embodiment of the present invention is described in detail with reference to the vibration data measured on the experimental test platform. A local defect is machined on the outer ring of the support bearing to simulate a fault. The method for diagnosing a weak fault of a pump bearing based on time-frequency adaptive segmentation provided by the embodiment of the present invention is executed. The bearing model is ER-16K, and the characteristic frequency of the bearing outer ring fault is about 80 Hz. The execution process includes the following operations:

[0128] Signal processing step: transforming and processing the collected vibration signal to obtain the corresponding time-frequency distribution characteristics;

[0129] Feature division step: quantify the adjacent correlation distribution relationship that characterizes the time-frequency distribution characteristics, realize frequency domain interval division, and obtain multi-part time-frequency distribution characteristics;

[0130] Sensitive frequency band determination steps: Use principal component analysis technology to determine the fault characteristic frequency ratio index corresponding to the time-frequency distribution characteristics of different parts, and select sensitive components;

[0131] Adaptive optimization processing step: based on the fault characteristic frequency ratio index, a differential search algorithm is used for optimization processing to obtain a fault characteristic component, and based on the fault characteristic component, bearing weak fault feature extraction and diagnosis of a machine pump are realized.

[0132] Wherein, (1) when collecting the vibration signal, the DT9837 data acquisition card is used to collect the vibration signal at the support bearing, the sampling frequency is 20480Hz, and the sampling time length is 1s. Figure 2 As shown in the figure, according to the current information, no obvious periodic impact can be observed from the figure, the bearing fault characteristic information is covered by the vibration signal of other components and noise, and further processing and mining are needed.

[0133] (2) Further time-frequency transformation, in order to extract the bearing fault characteristic, the vibration signal is mapped to the time-frequency domain for processing through short-time Fourier transform, in order to make the bearing periodic impact information more obvious, according to the foregoing sampling frequency and bearing structure, the window length of 300 points is selected by the application, and the time-frequency distribution of the vibration signal is as shown in Figure 3 At this time, the time-frequency distribution is still relatively complex and needs further processing and analysis.

[0134] (3) The correlation degree of adjacent components in the frequency dimension is calculated according to the following formula,

[0135]

[0136] Wherein, STFT(τ,f) and STFT(τ,f+1) represent the time-frequency distribution of adjacent frequency axes, and if the similarity of the distribution of adjacent frequency axes is higher, the value of Corr(f) is larger. The time-frequency adjacent correlation distribution of the vibration signal is as shown in Figure 4 From which six local maxima can be observed.

[0137] (4) Time-frequency boundary division:

[0138] The local minimum value between the two local maximum values of the time-frequency distribution adjacent correlation is taken as the boundary, and the signal time-frequency distribution is divided into six parts n =[ω n-1 ,ω n ], n = 1, 2, …, N.

[0139] (5) Fault sensitive frequency band selection

[0140] The principal component analysis method is used to obtain the principal component of each part and obtain the corresponding frequency spectrum, and the essence of principal component analysis is to determine the corresponding coefficient according to the mutual relationship of multi-dimensional components and then combine them into low-dimensional components.

[0141] The principal component P of each time-frequency part is obtained by using the principal component analysis technology:

[0142] P(t, n) = PCA[STFT(t, Λ n )] n = 1,2, …, N

[0143] Then the spectrum of each principal component is calculated respectively,

[0144]

[0145] The fault feature information contained in each time-frequency part is measured by the fault frequency ratio index, and the time-frequency component with the maximum fault feature frequency ratio index is selected as the sensitive component for further processing.

[0146]

[0147] The distribution of the fault frequency ratio index of the analyzed signal is shown in Figure 5 The sixth part with the highest fault frequency ratio index of the principal component spectrum is defined as the fault sensitive frequency band. The sixth part corresponding to the principal component spectrum is shown in Figure 6 From which the bearing inner ring fault feature frequency of 78 Hz can be observed.

[0148] (6) Frequency coefficient adaptive optimization based on differential search algorithm

[0149] In order to further enhance the bearing fault feature and eliminate the influence of other components in the gearbox, the differential search algorithm is introduced to perform adaptive optimization on the coefficients of the frequency axis of the sixth part with the fault frequency ratio index as the fitness function. The specific steps are as follows:

[0150] ① Determine the number of population individuals N and the maximum optimization iteration number G;

[0151] ② Initialize the position of the individual using the following formula, the optimization coefficient range is [0, 1], and the fitness of the current position of the individual is calculated according to the fitness function device fault frequency ratio index;

[0152] x i,j =rand(up j -low j )+low j

[0153] ③ Calculate the temporary residence position of all individuals according to the following formula, if the temporary residence position exceeds the boundary [0, 1] of the optimization space, move it to a random position within the optimization space;

[0154] S=X i +γ·(donor-X i )

[0155] IV. Comparing the fitness function of the temporary stay position and the current position, judging whether the fitness of the temporary stay position is better than the current position or not, if the temporary stay position is better than the current position, the individual immediately moves to the temporary stay position and updates the fitness, otherwise, staying in the original position, after completion, the individual waits for other individuals in the population to complete the step, when all individuals complete the exploration and comparison of the temporary stay position, the population and the global optimal position are updated;

[0156]

[0157] V. If the total number of iterations is not reached, return to step III, if the total number of iterations is reached, output the global optimal position and the fitness.

[0158] After obtaining the coefficient, the sixth part of each frequency axis information is weighted and combined, and the combined component spectrum is as shown in Figure 7 Compared with the principal component spectrum as shown in Figure 6 The bearing fault feature is obviously enhanced.

[0159] As can be seen, the weak fault diagnosis method and system for machine pump bearing based on time-frequency adaptive segmentation provided by the application transform and process the collected vibration signals to obtain corresponding time-frequency distribution characteristics, quantize the adjacent correlation distribution relationship of the time-frequency distribution characteristics, realize frequency domain interval division to obtain multiple time-frequency distribution characteristics, the embodiment of the application considers the joint distribution of the vibration signal in time domain-frequency domain, the analysis of the vibration signal is more comprehensive, the relationship between each frequency band is judged by using the distribution manifold of the vibration signal in time-frequency domain, adaptive decomposition of different frequency band signals is realized, and the decomposition result is more reasonable.

[0160] Further, the principal component analysis technology is used to determine the fault characteristic frequency ratio index corresponding to the different part time-frequency distribution characteristics, and the sensitive component is selected, and then the differential search algorithm is used for coefficient optimization processing, the fault characteristic component is calculated and obtained, and the weak fault characteristic extraction and diagnosis of the machine pump bearing are realized based on the same, the fault sensitive frequency band is selected by using the fault characteristic frequency ratio index of the principal component spectrum of each part obtained by decomposition, and on this basis, the adaptive optimization of the frequency axis coefficient of the sensitive frequency band is combined with the differential search algorithm, the bearing fault feature is further enhanced, the influence of other components in the gearbox is eliminated, accurate weak fault diagnosis is realized, the operation process is simple, and the timeliness is good, the predictive maintenance of the moving equipment is guided, and long-term stable operation of the device and safe industrial production are ensured.

[0161] For each of the foregoing method embodiments, for the sake of simple description, each is described as a combination of a series of acts, but those skilled in the art should appreciate that the present application is not limited by the order of the described acts, because, according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should appreciate that the embodiments described in the specification are all preferred embodiments, and the acts and modules involved are not necessarily essential to the present application.

[0162] It should be noted that in other embodiments of the present application, the method can also be obtained by combining one or more of the above embodiments to obtain a new machine pump bearing weak fault diagnosis method based on time-frequency adaptive segmentation, so as to realize the timely discovery and relief of the weak fault of the industrial machine pump bearing.

[0163] It should be noted that based on the method in any one or more of the above embodiments of the present application, the present application also provides a storage medium having a program code for implementing the method in any one or more of the above embodiments stored thereon, and the code can implement the machine pump bearing weak fault diagnosis method based on time-frequency adaptive segmentation when executed by an operating system.

[0164] Embodiment two:

[0165] The above embodiments of the present application disclose the method in detail, and the method of the present application can be implemented in various forms of devices or systems, so based on other aspects of the method in any one or more of the above embodiments, the present application also provides a machine pump bearing weak fault diagnosis system based on time-frequency adaptive segmentation, which is used to execute the machine pump bearing weak fault diagnosis method based on time-frequency adaptive segmentation in any one or more of the above embodiments. The following specific embodiments are given for detailed description.

[0166] Specifically, the machine pump bearing weak fault diagnosis system based on time-frequency adaptive segmentation provided in the embodiments of the present application comprises:

[0167] A signal processing module configured to transform and process the collected vibration signal to obtain corresponding time-frequency distribution characteristics;

[0168] A feature division module configured to quantize the adjacent correlation distribution relationship representing the time-frequency distribution characteristics, realize frequency domain interval division, and obtain time-frequency distribution characteristics in a multi-part form;

[0169] A sensitive frequency band determination module configured to determine the fault feature frequency ratio index corresponding to different part time-frequency distribution characteristics by using principal component analysis technology, and select sensitive components;

[0170] An adaptive optimization processing module is configured to perform optimization processing based on the fault feature frequency ratio index by using a differential search algorithm to obtain a fault feature component, and based on which to realize weak fault feature extraction and diagnosis of the machine pump bearing.

[0171] Optionally, in an embodiment, the signal processing module calculates a corresponding time-frequency distribution feature based on the collected vibration signal of the machine pump bearing by using short-time Fourier transform.

[0172] Further, in an embodiment, the signal processing module calculates a corresponding time-frequency distribution feature of the vibration signal according to the following formula:

[0173]

[0174] In the formula, STFT(t, f) represents the time-frequency distribution feature of the signal, f represents the frequency axis, s(τ) represents the vibration signal of the machine pump device, h(τ-t) represents a window function for segmenting the input signal for short-time Fourier transform, and τ represents a time parameter inside the window function.

[0175] In a preferred embodiment, the signal processing module uses a window function with fixed time-frequency shape in the process of calculating the corresponding time-frequency distribution feature, and the resolution in the time-frequency domain does not change with time and frequency, and the corresponding time-frequency grid size is fixed.

[0176] In an embodiment, the feature division module quantifies the adjacent correlation distribution relationship representing the time-frequency distribution feature by calculating the distribution similarity of adjacent components in the frequency dimension.

[0177] Further, in an embodiment, the feature division module calculates the distribution similarity of adjacent components in the frequency dimension according to the following formula:

[0178]

[0179] In the formula, STFT(τ, f) and STFT(τ, f+1) represent the time-frequency distribution of two adjacent frequency axes f and f+1, respectively, and Corr(f) represents the adjacent correlation distribution similarity of the time-frequency distribution feature.

[0180] Optionally, in an embodiment, the feature division module is configured to use the midpoint between two adjacent maximum points of the adjacent correlation distribution similarity as a division boundary to adaptively decompose the frequency domain support interval of the signal to be processed into multiple parts to form a time-frequency distribution feature in the form of multiple parts.

[0181] In a preferred embodiment, the sensitive frequency band determination module performs the following operations:

[0182] The principal components of each time-frequency part are calculated by using principal component analysis technology, and the spectrum of the principal components of each time-frequency part is obtained.

[0183] The fault characteristic information contained in each time-frequency part is measured by a fault frequency ratio index, and a time-frequency component with the maximum fault characteristic frequency ratio index is selected as a fault sensitive component.

[0184] Further, in an embodiment, the sensitive frequency band determination module logically calculates the principal components of each time-frequency part according to the following formula:

[0185] P(t, n) = PCA[STFT(t, Λ n )] n = 1, 2, …, N

[0186] wherein P(t, n) represents the principal component corresponding to the nth part time-frequency signal, STFT(t, Λ n ) represents the nth part time-frequency signal, and N represents the total number of N part time-frequency signals.

[0187] Optionally, in an embodiment, the adaptive optimization processing module performs adaptive optimization on the coefficients of each frequency axis of the sensitive part by taking the obtained fault characteristic frequency ratio index as an adaptive function.

[0188] Further, in an embodiment, the adaptive optimization processing module performs adaptive optimization according to the following operations:

[0189] determining the number of population individuals and the maximum optimization iteration number;

[0190] initializing the position of the individual, and calculating the fitness of the current position of the individual according to the fitness function;

[0191] calculating the temporary stop position of all individuals, and if the temporary stop position exceeds the boundary of the optimization space, moving the temporary stop position to a random position within the optimization space, and calculating the corresponding fitness;

[0192] comparing the fitness of the temporary stop position of the individual with the fitness of the current position of the individual, and if the former is better, moving the position of the individual to the temporary stop position and updating the fitness, otherwise keeping the original position;

[0193] after all individuals perform the deletion operation, updating the population and the global optimal position; and performing a loop operation until a set requirement is met to output the global optimal position and the fitness thereof, and obtaining the corresponding optimization coefficient as the target coefficient.

[0194] In the weak fault diagnosis system for a pump bearing based on time-frequency adaptive segmentation provided by the embodiment, each module or unit structure can be independently operated or combinedly operated according to actual signal processing requirements and index calculation requirements, so as to realize corresponding technical effects.

[0195] It is to be understood that the embodiments disclosed herein are not limited to particular structures, processes, or materials disclosed herein but are extended to equivalents thereof The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting.

[0196] It is to be understood that the embodiments disclosed herein are not limited to particular structures, processes, or materials disclosed herein but are extended to equivalents thereof The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting.

[0197] Although the present application has been described with reference to specific embodiments, it will be apparent to those skilled in the art that various modifications and changes can be made without departing from the spirit and scope of the application. Accordingly, the specification and figures are to be regarded in an illustrative manner and the scope of the application is to be indicated by the appended claims.

Claims

1. A method for diagnosing weak faults of pump bearings based on time-frequency adaptive segmentation, characterized in that: The method comprises: Signal processing step: transforming and processing the collected vibration signal to obtain the corresponding time-frequency distribution characteristics; Feature division step: quantify the adjacent correlation distribution relationship that characterizes the time-frequency distribution characteristics, realize frequency domain interval division, and obtain multi-part time-frequency distribution characteristics; Sensitive frequency band determination steps: Use principal component analysis technology to determine the fault characteristic frequency ratio index corresponding to the time-frequency distribution characteristics of different parts, and select sensitive components; Adaptive optimization processing steps: Based on the fault feature frequency ratio index, a differential search algorithm is used to perform optimization processing to obtain the fault feature component, and based on it, the weak fault feature extraction and diagnosis of the machine pump bearing is realized.

2. The method according to claim 1, characterized in that In the signal processing step, short-time Fourier transform is used to calculate the corresponding time-frequency distribution characteristics based on the collected pump bearing vibration signal.

3. The method according to claim 1 or 2, characterized in that In the signal processing step, the time-frequency distribution characteristics corresponding to the vibration signal are calculated according to the following logic: Where STFT(t,f) represents the time-frequency distribution characteristics of the signal, f represents the frequency axis, s(τ) represents the vibration signal of the pump equipment, h(τ-t) represents the window function that divides the input signal into segments for short-time Fourier transform, and τ represents the time parameter inside the window function.

4. The method according to claim 1, wherein In the signal processing step, when calculating the corresponding time-frequency distribution characteristics, a window function with a fixed time-frequency shape is used. The resolution in the time-frequency domain does not change with time and frequency, and the corresponding time-frequency grid size is fixed.

5. The method according to claim 1, wherein In the feature partitioning step, the distribution similarity of adjacent components in the frequency dimension is calculated to quantify the adjacent correlation distribution relationship that characterizes the time-frequency distribution features.

6. The method according to claim 5, characterized in that The distribution similarity of adjacent components in the frequency dimension is calculated as follows: Among them, STFT(τ,f) and STFT(τ,f+1) represent the time-frequency distribution of two adjacent frequency axes, frequency axis f and frequency axis f+1, respectively, and Corr(f) represents the similarity of adjacent correlation distribution of time-frequency distribution features.

7. The method according to claim 1, characterized in that In the feature division step, the midpoint between two adjacent maximum points of adjacent correlation distribution similarity is used as the division boundary to adaptively decompose the frequency domain support interval of the signal to be processed into multiple parts to form a multi-part time-frequency distribution feature.

8. The method according to claim 1, characterized in that The sensitive frequency band determination step includes the following operations: The principal component analysis technique is used to obtain the principal components of each time-frequency part and obtain the spectrum of the principal components of each time-frequency part; The fault frequency ratio index is used to measure the fault feature information contained in each time-frequency part, and the time-frequency component with the largest fault feature frequency ratio index is selected as the fault sensitive component.

9. The method according to claim 8, characterized in that Calculate the principal components of each time-frequency part according to the following logic: P(t,n)=PCA[STFT(t,Λ n )]n=1,2,…,N Among them, P(t,n) represents the principal component corresponding to the nth part of the time-frequency signal, STFT(t,Λ n ) represents the time-frequency signal of the nth part, and N represents the time-frequency signal of N parts in total.

10. The method according to claim 1, characterized in that In the adaptive optimization processing step, the coefficients of each frequency axis of the sensitive part are adaptively optimized using the obtained fault characteristic frequency ratio index as the fitness function.

11. The method according to claim 1 or 10, characterized in that The process of adaptive optimization includes: Determine the number of individuals in the population and the maximum number of optimization iterations; Initialize the individual's position and calculate the fitness of the individual's current position according to the fitness function; Calculate the temporary stay position of all individuals. If they exceed the limit of the optimization space, move them to a random position in the optimization space and calculate the corresponding fitness. Compare the fitness of the individual's temporary stay position with the fitness of the current position. If the former is better, move the individual's position to the temporary stay position and update the fitness; otherwise, keep the original position. After all individuals have been deleted, the population and the global optimal position are updated; the operation is repeated until the global optimal position and its fitness are output as required, and the corresponding optimization coefficient is obtained as the target coefficient.

12. A storage medium, characterized in that: The storage medium stores program code that can implement the method according to any one of claims 1 to 11.

13. A pump bearing weak fault diagnosis system based on time-frequency adaptive segmentation, characterized by: The system executes the method according to any one of claims 1 to 11.

Images