CN113567123B - Automatic diagnosis method for impact faults of rotary machinery - Google Patents
Automatic diagnosis method for impact faults of rotary machinery Download PDFInfo
- Publication number
- CN113567123B CN113567123B CN202110594020.6A CN202110594020A CN113567123B CN 113567123 B CN113567123 B CN 113567123B CN 202110594020 A CN202110594020 A CN 202110594020A CN 113567123 B CN113567123 B CN 113567123B
- Authority
- CN
- China
- Prior art keywords
- impact
- data
- frequency
- fault
- spectrum
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses an automatic diagnosis method for impact faults of rotary machinery, which comprises four parts, namely, determination of sampling parameters, data acquisition, extraction of impact spectrum and fault diagnosis, wherein the four parts are used for realizing automatic diagnosis for the impact faults of the rotary machinery, screening out the impact faults through a natural frequency criterion, calculating a center frequency and an analysis bandwidth, calculating the sampling number based on the parameters, calculating the sampling rate suitable for the characteristics of the rotary machinery, acquiring data, filtering, FFT (fast Fourier transform) and extraction of the impact spectrum from the acquired data, and realizing qualitative, quantitative, positioning analysis and diagnosis for the impact faults through the impact fault diagnosis method based on the impact general data. The invention has the beneficial effects that: the problem of single application scene of fault diagnosis of the rotary mechanical impact type caused by hardware resonance demodulation is avoided, the application scene of fault diagnosis of the impact type is widened, the design cost of the system is reduced, and the use flexibility of the diagnosis system is improved.
Description
Technical Field
The invention belongs to a fault feature extraction method of rotary machinery, in particular to an automatic diagnosis method of impact faults of rotary machinery.
Background
The resonance demodulation technique (DRT-Demodulated Resonance Technique) is also called an early fault detection method (IFD-Incipient Failure Detection), and is a rotating machinery fault diagnosis method widely applied at present. The vibration attenuation caused by the fault impact of the rotary machine is amplified by using the resonance of a measured piece, a sensor or a circuit, so that the sensitivity of fault monitoring is improved; meanwhile, the rotary machine fault information is extracted by using a demodulation technology, and the rotary machine fault is diagnosed by carrying out spectrum analysis on the demodulated information. Resonance demodulation technology was first developed in 1974, and the united states boeing company invented this patent, called the "resonance demodulation analysis system", which is the precursor of the current resonance demodulation technology; the technology amplifies and separates fault signals, improves the signal to noise ratio of the signals, and can embody the advantages of the method in early stage of bearing faults. Compared with the common spectrum analysis method and the impact pulse technology, the common spectrum analysis method and the impact pulse technology have the characteristic that the impact pulse is caused by a fault point, and the progress is that the resonance demodulation method can diagnose the fault position and the severity, so that the method is widely used once being developed, and plays an irreplaceable role in a plurality of fields until now.
The traditional resonance demodulator is used for detecting impact information generated by abnormal collision of a rotary machine by using sensing technologies such as piezoelectricity, strain and the like, amplifying weak impact signals into resonance waveforms of high-frequency free damping oscillations through resonance response of the resonators by using electric, mechanical or other (surface acoustic wave, acoustic and the like) hardware resonators, demodulating the resonance waveforms to obtain impact envelope signals without low-frequency vibration interference, and finally extracting the information of low-frequency impact through frequency spectrum analysis of the impact envelope signals to judge the existence and degree of faults of the rotary machine; the process is shown in figure 2. The conventional resonance demodulator has the following disadvantages in use: (1) The high-frequency resonance frequency of the tested rotary machine must be determined in advance through an impact test to determine the center frequency of the resonance band-pass filter; in practical engineering, the selection of the resonant bandpass filter has a decisive influence on the final diagnosis result, the poor design of the resonant bandpass filter will easily lead to diagnosis or missed diagnosis, and in order to determine the high-frequency resonance frequency of the rotating machine, an impact test is generally required, but for most users, the impact test cannot be performed due to the existence of various objective factors, and the center frequency of the resonant bandpass filter is generally set empirically. (2) The center frequency and bandwidth of the resonance band-pass filter are generally fixed, the positions of the high-frequency natural vibration of the resonance band-pass filter are different for different rotary machines, and the fault characteristic frequencies are also different, so that the designed resonance demodulator can be disabled for different bearing systems by adopting the fixed center frequency and bandwidth. (3) For an electric resonance demodulator, an active RC filter circuit is generally adopted as a resonance band-pass filter, the center frequency and the bandwidth of the resonance band-pass filter are set through discrete resistors and capacitors, the resonance band-pass filter is greatly influenced by ambient temperature and humidity, and the resonance band-pass filter is difficult to debug and poor in product consistency and is not suitable for mass production and popularization.
Disclosure of Invention
In order to overcome the technical problems, the invention discloses an automatic diagnosis method for impact faults of rotary machinery.
The technical scheme of the invention is as follows: an automatic diagnosis method for impact faults of rotary machinery comprises four parts including sampling parameter determination, data acquisition, impact spectrum extraction and fault diagnosis, and is used for realizing the known axial frequencyf n Natural frequency of sensorf c1 Automatic diagnosis of impact faults of the rotary machine with the highest fault characteristic value alpha of the tested piece; the sampling parameter determining part is carried out in three steps, wherein the first step is according to the sampling frequencyf s0 ≥2f c1 Sampling rotary machine monitoring locationx n The data are combined and subjected to an FFT,x n =2 a the method comprises the steps of a, classifying faults of a rotary mechanical monitoring part by using a natural frequency criterion, and analyzing a 3-order fault characteristic spectrum according to the frequency of a spectrum resonance band center maximum amplitude spectral line when an impact fault existsf c Determining an analysis bandwidth asf c -3.1αf n ~f c +3.1αf n The method comprises the steps of carrying out a first treatment on the surface of the The data acquisition part is performed in four steps, wherein the first step is to preset the sampling frequencyf s1 ≥2(f c +3.1αf n ) An analysis window width of 2 M And combining the analysis bandwidth to obtain the required sampling data number Ñ =INT [ (2) M-1 f s1 )/(3.1αf n )+0.5]Second step q=2 M+1 P=int (Ñ/q+0.5), actual sampling data point n=p×q, and adjusting sampling frequencyf s ＝12.4αf n P, the third step repeatedly executes the first and second steps until the sampling frequencyf s ≥2(f c +3.1αf n ) Fourth step according to sampling frequencyf s Sampling N data and storing the data in P groups of Q points; the impact spectrum extraction part is carried out in six steps, a data pointer K=0 is arranged after the data acquisition is finished in the first step, 1 group of Q point acquisition data is taken according to the data pointer K and assigned to K by the data pointer K+1 in the second step, the Q point data is subjected to FIR passband filtering and then is subjected to FFT and cached in the third step, the second step and the third step are repeatedly executed until the data pointer K=P in the fourth step, and the resonance center frequency is calculated in the fifth stepf c Spectral line position L c ＝INT(N*f c /f s +0.5), then from L c -2 M Start calculation at 2 M+1 DFT spectral line, sixth step of spectral line L c And the amplitude of each of the front and rear 3 spectral lines is 0 and the spectral line L is recorded c The zero frequency is 2 before and after M The spectrum lines are symmetrically added to obtain an impact spectrum; the fault diagnosis part performs qualitative, quantitative and positioning analysis on the faults according to the vibration or impact fault diagnosis method according to the different fault classifications of the second step of the sampling parameter determination part, and automatically jumps to the first step of the sampling parameter determination part if the fault monitoring is required to be continued, and otherwise exits from the fault automatic diagnosis cycle.
In the invention, the natural frequency criterion is pairx n The data of the individual FFTs are subjected to a discretization decision,x n =2 a a is a positive integer, and the characteristic spectrum used for distinguishing the impact faults and other various interference spectrums are determined discretely by FFTx n Searching and sensor natural frequency in dataf n Equal or similar frequency values, then making extreme point judgment, if the extreme point is calculatedx n The average amplitude of 2 spectral lines is compared with the amplitude corresponding to the spectral line of the extreme point, the average amplitude is smaller than the amplitude corresponding to the spectral line of the extreme point, and the amplitude of the spectral line of the extreme point is the frontAnd 30 maximum extreme points, the spectral lines are discrete, and the impact type faults can be judged.
In the invention, the FIR passband filtering method is to take 1 group of Q point sampling data from P groups of data in sequence, respectively supplement K/2 zeros before and after the sampling data, and take K data and filtering coefficients in sequence in Q+K point data each timeh(k) And multiplying, accumulating and summing, and averaging the summed value, wherein the averaged value is the filtered value.
In the invention, the DFT spectral line calculation method is to obtain the center frequencyf c Spectral line position L c Calculating the initial position of the required analysis frequency spectrum as L c -2 M Analyzing the data position of the spectrum start position in the Q-point FFT resultl ` =L c -2 M -Q*INT((L c -2 M ) Q) for use in P sets of Q-point FFT resultsl ` The position data is used for obtaining 1 refinement spectrum value according to the DFT principle.
In the present invention, the impact spectrum extraction method is divided into pair determination L c Will L c The amplitude of each 3 spectral lines is 0, and the spectral line L is calibrated c The corresponding frequency is the zero frequency position of the frequency spectrum coordinate system, and the spectral line L c The spectral lines on the left and right sides are mutually called as spectral lines and are added to obtain an impact spectrum.
In the invention, the impact fault diagnosis method is to extract the impact frequency by an impact spectrum extraction method, and effectively remove various interferences by respectively using a fault multi-order criterion, a fault spectral line discreteness criterion and a fault characteristic spectral line energy factor criterion on the impact spectrum, so as to realize accurate identification of the impact faults.
The beneficial effects of the invention are as follows: the method comprises the steps of screening out impact faults through natural frequency criteria, determining the center frequency and the analysis bandwidth of a resonance frequency band, obtaining the number of sampling data according to the center frequency and the analysis bandwidth, pushing out the sampling frequency suitable for the parameter characteristics of the rotary machine according to the number of the sampling data, collecting data, performing FIR band-pass filtering, FFT and DFT spectral line analysis on the collected data to obtain an impact spectrum, and finally realizing qualitative, quantitative and positioning analysis on the rotary machine by an impact fault diagnosis method based on the impact spectrum, so that the single problem of an application scene of the diagnosis on the rotary machine impact faults caused by hardware resonance demodulation is avoided, the application scene of the diagnosis on the rotary machine impact faults is widened, the design cost of a system is reduced, and the use flexibility of a diagnosis system is improved.
Drawings
FIG. 1 is a flow chart of a method for automatically diagnosing a rotary machine impact type fault according to the present invention;
fig. 2 is a hardware resonance demodulator workflow.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention; it will be apparent that the described embodiments are only some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, the invention is a flow chart of an automatic diagnosis method for impact faults of a rotary machine, which comprises four parts of sampling parameter determination, data acquisition, impact spectrum extraction and fault diagnosis, and is used for realizing the known axial frequencyf n Natural frequency of sensorf c1 Automatic diagnosis of impact faults of the rotary machine with the highest fault characteristic value alpha of the tested piece; the sampling parameter determining part is carried out in three steps, wherein the first step is according to the sampling frequencyf s0 ≥2f c1 Sampling rotary machine monitoring locationx n The data are combined and subjected to an FFT,x n =2 a the method comprises the steps of a, classifying faults of a rotary mechanical monitoring part by using a natural frequency criterion, and analyzing a 3-order fault characteristic spectrum according to the frequency of a spectrum resonance band center maximum amplitude spectral line when an impact fault existsf c Determining an analysis bandwidth asf c -3.1αf n ~f c +3.1αf n The method comprises the steps of carrying out a first treatment on the surface of the The data acquisition part is performed in four steps, wherein the first step is to preset the sampling frequencyf s1 ≥2(f c +3.1αf n ) An analysis window width of 2 M And combining the analysis bandwidth to obtain the required sampling data number Ñ =INT [ (2) M-1 f s1 )/(3.1αf n )+0.5]Second step q=2 M+1 P=int (Ñ/q+0.5), actual sampling data point n=p×q, and adjusting sampling frequencyf s ＝12.4αf n P, the third step repeatedly executes the first and second steps until the sampling frequencyf s ≥2(f c +3.1αf n ) Fourth step according to sampling frequencyf s Sampling N data and storing the data in P groups of Q points; the impact spectrum extraction part is carried out in six steps, a data pointer K=0 is arranged after the data acquisition is finished in the first step, 1 group of Q point acquisition data is taken according to the data pointer K and assigned to K by the data pointer K+1 in the second step, the Q point data is subjected to FIR passband filtering and then is subjected to FFT and cached in the third step, the second step and the third step are repeatedly executed until the data pointer K=P in the fourth step, and the resonance center frequency is calculated in the fifth stepf c Spectral line position L c ＝INT(N*f c /f s +0.5), then from L c -2 M Start calculation at 2 M+1 DFT spectral line, sixth step of spectral line L c And the amplitude of each of the front and rear 3 spectral lines is 0 and the spectral line L is recorded c The zero frequency is 2 before and after M The spectrum lines are symmetrically added to obtain an impact spectrum; the fault diagnosis part performs qualitative, quantitative and positioning analysis on the faults according to the vibration or impact fault diagnosis method according to the different fault classifications of the second step of the sampling parameter determination part, and automatically jumps to the first step of the sampling parameter determination part if the fault monitoring is required to be continued, and otherwise exits from the fault automatic diagnosis cycle.
The specific method comprises the following steps: the method comprises the steps of determining a sampling parameter part, acquiring operation data of a rotary machine by using a multi-parameter sensor, processing and analyzing the data, screening out impact faults according to a natural frequency criterion, and simultaneously obtaining the center frequency and the analysis bandwidth of a resonance frequency band according to the general fault characteristics of the rotary machine, wherein the three steps are executed:
the first step is to preset sampling rate during system initializationf s0 According to sampling frequencyRate off s0 Samplingx n The rotating machine monitors the site data and performs an FFT,x n =2 a a is a positive integer, where requirements aref s0 ≥2f c1 To avoid frequency aliasing caused by too low sampling rate in data analysis,f c1 for the natural frequency of the sensor, for the configuration parameters of the selected sensor, the number of samples isx n Based on natural frequency of sensorf c1 It is determined that the number of the cells,x n =2 a a is a positive integer; the second step is to reject the interference signal of the frequency spectrum data obtained after FFT, then to classify the fault of the monitoring part of the rotating machine by using the natural frequency criterion, to execute the impact fault diagnosis method for impact faults and to execute the vibration fault diagnosis method for vibration faults, wherein the natural frequency criterion is the pair of the natural frequency criterionx n The data of the FFT are subjected to discreteness judgment, because the impact generated by faults of the rotating machine in the operation process is periodic, namely the fault impact always generates neither in advance nor behind according to a fixed time interval period, corresponding fault spectral lines in an impact spectrum caused by the faults of the rotating machine have better discreteness, namely isolated fault characteristic spectral lines protrude from noise, and the distinction between system white noise, locomotive conventional vibration noise, continuous spectrum formed by single interference pulse or other fault-like spectral lines and impact fault characteristic spectrum in the collected data can be well realized, and the implementation process of the discreteness judgment method is as follows: from FFTx n Searching and sensor natural frequency in dataf n Equal or similar frequency values, judging extreme points, calculating the extreme pointsx n The average amplitude of 2 spectral lines is compared with the amplitude corresponding to the spectral line of the extreme point, the average amplitude is smaller than the amplitude corresponding to the spectral line of the extreme point, and meanwhile, the amplitude of the spectral line of the extreme point is the first 30 maximum extreme points, so that the spectral line is discrete, meanwhile, the impact fault can be judged, after the acquired data is subjected to the natural frequency criterion as the impact fault, the third step is executed downwards, and when the impact fault exists, the third step is carried out according to the rotating machineGeneral fault characteristics of the machine are considered and analyzed to obtain a 3-order fault characteristic spectrum according to the frequency of spectral line with the maximum amplitude in the center of a frequency spectrum resonance bandf c Determining an analysis bandwidth asf c -3.1αf n ~f c +3.1αf n Whereinf c The impact faults can be determined simultaneously when the natural frequency criterion is adopted to judge the impact faults, alpha is the highest fault characteristic value of the tested piece, and is determined according to the parameters of the rotary machine,f n is the axial frequency of the rotating machine during operation.
The data acquisition part is used for acquiring the data according to the center frequency of the frequency bandf c Analysis of bandwidthf c -3.1αf n ~f c +3.1αf n In combination with a predetermined sampling frequencyf s1 An analysis window width of 2 M Acquiring the number of sampling data of monitoring points of the rotary machine, reversely deducing the sampling frequency suitable for the characteristics of the rotary machine based on the number of sampling data to obtain the optimal sampling frequency under different parameters of the characteristics of the rotary machine, and providing a data basis for impact general calculation, wherein the specific flow comprises four steps:
first step, presetting an initial sampling frequencyf s1 Determining the width of the analysis window to be 2 according to the FFT calculation principle M By analysing bandwidthf c -3.1αf n ~f c +3.1αf n Determining the signal width to be 2 multiplied by 3.1 alphaf n Resolution is deltaf=2×3.1αf n /2 M Theoretical sampling point Ñ =calculated from sampling frequency and resolutionf s1 /ΔfCombined signal width 2×3.1αf n Sampling ratef s1 Window width determines the number of sample data to be Ñ =int [ (2) M-1 f s1 )/(3.1αf n )+0.5]Wherein the acquisition of the sampled data can cover an entire analysis bandwidth and the acquired signals do not suffer from aliasing requirementsf s1 ≥2(f c +3.1αf n ) The method comprises the steps of carrying out a first treatment on the surface of the The second step is based on the FFT symmetry principle q=2 x Then 2 can be observed x-1 Each group of data is Q point according to any effective spectral lineThe base FFT adopts the principle of a feed method to divide data into groups P=INT (Ñ/Q+0.5), the number of the sampled data to be processed actually is N=P×Q, the grouping P can be increased as much as possible according to the actual requirement, and the Q point is reduced as much as possible, so that Q=2 is taken M+1 P=int (Ñ/q+0.5), taking the actual sampling point number as n=p×q, and re-determining the sampling frequency according to the actual sampling point numberf s ＝12.4αf n P is as follows; third step for the method determined in the second stepf s Make a judgment iff s ≥2(f c +3.1αf n ) Then consider the determination in the second stepf s The sampling frequency meets the requirement of sampling the data of the monitoring point of the rotary machine, otherwise, the first step is returned to, and the first step and the second step are re-executed until the sampling frequencyf s ≥2(f c +3.1αf n ) Until that is reached; fourth step according to sampling frequencyf s N data points are sampled, and the data in the sampling process is stored in P groups of Q points.
The impact spectrum extraction part is used for carrying out FIR band-pass filtering, FFT and DFT spectral line obtaining on the data acquired by data acquisition to acquire an impact spectrum, providing a data base for the next impact diagnosis method, and specifically carrying out the steps of:
the first step is to set a data pointer K=0 after the data acquisition is finished, so as to avoid confusion caused by extracting and processing the data of the P group of Q points, thereby influencing the diagnosis result; step two, extracting 1 group of Q point data in the P groups of data according to the value of the data pointer K to process the subsequent process, adding 1 to the data pointer K, and assigning the K+1 value to the K value to realize acquisition data according to the data sequence stored during data acquisition and process the subsequent process; thirdly, performing FIR passband filtering on the extracted Q point data, wherein the purpose of the FIR passband filtering is to filter out low-frequency signals and high-frequency signals acquired by the data, and the implementation process of the FIR passband filtering method is as follows: first according to the analysis bandwidthf c -3.1αf n ~f c +3.1αf n And sampling frequencyf s Offline computing and caching filter coefficients of order Kh(k) Since the FFT point number is Q, band-pass filtering is based on analysisThe wave complex modulation refined spectrum analysis method can obtain the filter coefficienth(k) The real part and the imaginary part of (a) are respectively:
in the method, in the process of the invention,is the filter order;k=0, 1,2, …, Q-1; from the above, complex analytic band pass filter coefficientsh(k) The calculation of (2) is only related to the number Q of the filter order K, FFT, time seriesx(n) And filter coefficientsh(k) It has no direct relation, so it can be off-line calculated and buffered for standby before data sampling, then take 1 group of Q point sampling data from P groups of data, respectively supplement K/2 zeros before and after Q point data to make Q+K data, then take K data from Q+K data in sequence, and then make the taken data and the taken datah(k) Multiplying, accumulating, summing and averaging to obtain a filtering value, wherein the calculation formula is as follows:in the method, in the process of the invention,iin order to calculate the sequence number of the filtering value, finally, Q point FFT is carried out on the filtered Q point data and the filtered Q point data is cached, and an arbitrary base FFT algorithm with Q being an integer power of 2 is adopted to directly carry out FFT operation on the Q point data, so that the advantages of the Q point data and the Q point data are fully utilized and combined; the fourth step is to repeatedly execute the second and third steps until the data pointer K=P, namely, pass band filtering and FFT operation and buffering of P groups of data are completed; fifth step, calculating the resonance center frequencyf c Spectral line position L c ＝INT(N*f c /f s +0.5), from L c -2 M Start calculation at 2 M+1 DFT lines to enable resonance center frequencyf c The spectral line position is positioned at the right center of the new spectrogram, and the DFT spectral line calculation method comprises the following steps: obtaining center frequencyf c Spectral line position L c Calculating the initial position of the required analysis frequency spectrum as L c -2 M Analyzing the spectrum initial position in the Q point FFT resultData location of (2)l ` =L c -2 M -Q*INT((L c -2 M ) and/Q), obtaining 1 refined spectrum value of the position data in the P groups of Q point FFT results according to the DFT principle, wherein the calculation formula is as follows:in the middle ofl ` =L c -2 M -Q*INT((L c -2 M ) Q) and 0,1,2 …,2 M+1 ；i=0，1，2，…，P-1；Y i (l ` )=Y i (Z*Q+y)=Y i (y)，Y i (y) is a value in the P groups of FFT results, Z is a positive integer, and 0 is less than or equal toyQ-1 is not more than; sixth step of spectral line L c And the amplitude of each of the front and rear 3 spectral lines is 0 and the spectral line L is recorded c The zero frequency is 2 before and after M The impact spectrum is obtained by symmetrically adding the spectral lines, and the impact spectrum extraction method comprises the following steps: for L c And the amplitude of each of the front and rear 3 spectral lines is set to 0, and the spectral line L is calibrated c The corresponding frequency is the zero frequency position of the spectrum coordinate system, and the spectral line L is obtained c The left and right pairs are mutually called spectral lines are added.
The fault diagnosis part carries out qualitative, quantitative and positioning analysis on faults according to the fault classification difference of the second step of the sampling parameter determination part respectively according to a vibration or impact fault diagnosis method, if fault monitoring is needed to be continued, the fault diagnosis part automatically jumps to the first step of the sampling parameter determination part, otherwise, the fault diagnosis part exits from the fault automatic diagnosis cycle, wherein the impact fault diagnosis method is to extract the impact frequency through an impact spectrum extraction method, and can accurately identify impact faults by using a fault multi-order criterion, a fault spectral line discrete criterion and a fault characteristic spectral line energy factor criterion respectively on the impact frequency spectrum, wherein the fault multi-order criterion is used because the precision diagnosis technology changes a fault impact signal into a specific object attenuation waveform and carries out FFT conversion to diagnose the faults, and the fault diagnosis method is different from the general direct vibration analysis, one sinusoidal vibration only has a first-order frequency spectrum, an envelope demodulation signal caused by fault impact contains abundant high-order frequency spectrums, and the fault signals are modulated due to factors such as a sensor installation position, a rotating machine material, other vibration on the rotating machine, and the like, and the fault diagnosis method is effectively realized if the fault diagnosis method is needed because the error diagnosis is carried out because the fault signals which contain a large quantity of the components caused by modulation and the fault components are not adopted: if there is a spectrum line with a mutual magnitude difference of more than three orders corresponding to a certain type of faults on the spectrogram, the faults are possible to exist, so that multistage judgment is carried out on FFT spectrum lines during precise diagnosis and analysis; the purpose of implementing the discretization judgment of the fault characteristic spectral lines is to check whether the discretization of the increment lines reaches the standard after the envelope demodulation spectrum is searched to obtain the characteristic spectral lines corresponding to various faults, and the existence of the characteristic spectral lines which do not meet the standard is denied, so that the fault spectral lines have better discretization, namely, isolated spectral lines are highlighted from noise, and the implementation process of the discretization criterion of the fault spectral lines is as follows: if the characteristic spectral line corresponds to an extreme point, calculating the average spectral amplitude in a specified range, comparing the average amplitude with the amplitude of the fault characteristic spectral line, wherein the average amplitude is smaller than a certain proportion of the amplitude of the friction characteristic spectral line, and the amplitude of the fault characteristic spectral line is the first 30 maximum extreme points; the purpose of implementing the fault characteristic spectral line energy factor criterion is that for the envelope demodulation spectral diagram which is impossible to be matched with the theoretical model by computer software under normal conditions, meanwhile, if the brain analysis of a person is not needed, a certain common envelope frequency modulation spectrum diagram obtained by actual analysis is automatically compared with the envelope demodulation spectral diagram which is supposed to be possessed in theory by computer software, so that the state of the rotating machinery is difficult to diagnose, therefore, the information which can reflect the envelope demodulation spectral characteristics most easily is extracted, the computer can easily identify the actual state of a running part of the locomotive according to the information, and the practice proves that in the envelope demodulation signal spectrum of the rotating machinery fault, the characteristic is that the characteristic has multiple steps, the ratio of the sum of spectral line amplitudes of the first order, the second order, the third order and the like of the fault characteristic frequency to the average amplitude of the measured spectrum is defined as the fault characteristic spectral line energy factor, and when the fault characteristic spectral line energy factor is larger than a certain specific value, the condition is proved to have good repeatability and is mainly used for judging whether the fault exists or not.
The invention has the beneficial effects that: the method comprises the steps of taking the natural frequency of a sensor which is more than twice as the sampling rate to collect data and perform preliminary analysis, screening out impact faults through a natural frequency criterion, determining the center frequency and the analysis bandwidth of a resonance frequency band, obtaining the number of sampling data through the center frequency and the analysis bandwidth, deducing the sampling frequency suitable for the parameter characteristics of the rotary machine through the number of sampling data, collecting data, performing FIR band-pass filtering, FFT and DFT spectral line analysis on the collected data to obtain an impact spectrum, and finally performing qualitative, quantitative and positioning analysis on the rotary machine through an impact fault diagnosis method based on the impact spectrum, so that the problem of single application scene of the fault diagnosis of the rotary machine caused by hardware resonance demodulation is avoided, the use scene of the fault diagnosis of the rotary machine is widened, the design cost of the system is reduced, and the use flexibility of the diagnosis system is improved.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, alternatives, and improvements that fall within the spirit and scope of the invention.
Claims (6)
1. An automatic diagnosis method for impact faults of rotary machinery comprises four parts including sampling parameter determination, data acquisition, impact spectrum extraction and fault diagnosis, and is used for realizing the known axial frequencyf n Natural frequency of sensorf c1 Automatic diagnosis of impact faults of the rotary machine with the highest fault characteristic value alpha of the tested piece; the method is characterized in that: the sampling parameter determining part is carried out in three steps, wherein the first step is according to the sampling frequencyf s0 ≥2f c1 Sampling rotary machine monitoring partBit ofx n The data are combined and subjected to an FFT,x n =2 a a is a positive integer greater than zero, the second step uses natural frequency criterion to classify faults at the rotary mechanical monitoring part, and the third step considers and analyzes 3-order fault characteristic spectrum when impact faults exist, and the frequency of spectral lines with the maximum amplitude is calculated according to the center of a frequency spectrum resonance bandf c Determining an analysis bandwidth asf c -3.1αf n ~f c +3.1αf n The method comprises the steps of carrying out a first treatment on the surface of the The data acquisition part is performed in four steps, wherein the first step is to preset the sampling frequencyf s1 ≥2(f c +3.1αf n ) An analysis window width of 2 M And combining the analysis bandwidth to obtain the required sampling data number Ñ =INT [ (2) M-1 f s1 )/(3.1αf n )+0.5]Second step q=2 M+1 P=int (Ñ/q+0.5), actual sampling data point n=p×q, and adjusting sampling frequencyf s ＝12.4αf n P, the third step repeatedly executes the first and second steps until the sampling frequencyf s ≥2(f c +3.1αf n ) Fourth step according to sampling frequencyf s Sampling N data and storing the data in P groups of Q points; the impact spectrum extraction part is carried out in six steps, a data pointer K=0 is arranged after the data acquisition is finished in the first step, 1 group of Q point acquisition data is taken according to the data pointer K and assigned to K by the data pointer K+1 in the second step, the Q point data is subjected to FIR passband filtering and then is subjected to FFT and cached in the third step, the second step and the third step are repeatedly executed until the data pointer K=P in the fourth step, and the resonance center frequency is calculated in the fifth stepf c Spectral line position L c ＝INT(N*f c /f s +0.5), then from L c -2 M Start calculation at 2 M+1 DFT spectral line, sixth step of spectral line L c And the amplitude of each of the front and rear 3 spectral lines is 0 and the spectral line L is recorded c The zero frequency is 2 before and after M The spectrum lines are symmetrically added to obtain an impact spectrum; the fault diagnosis part performs qualitative, quantitative and positioning analysis on the faults according to the vibration or impact fault diagnosis method according to the difference of fault classifications in the second step of the sampling parameter determination part, and automatically jumps to the sampling parameter determination if the fault monitoring is required to be continuedAnd (3) determining the first step of the part, and otherwise, exiting the fault automatic diagnosis cycle.
2. The automatic diagnosis method for impact faults of rotary machines according to claim 1, characterized in that the natural frequency criterion is a pair ofx n The data of the individual FFTs are subjected to a discretization decision,x n =2 a a is a positive integer greater than zero, and the method for discretizing the characteristic spectrum of the impact type fault from other various interference spectrums is that the characteristic spectrum is obtained by FFTx n Searching and sensor natural frequency in dataf n Equal or similar frequency values, then making extreme point judgment, if the extreme point is calculatedx n And (2) comparing the average amplitude of the spectral lines with the amplitude corresponding to the spectral line of the extreme point, wherein the average amplitude is smaller than the amplitude corresponding to the spectral line of the extreme point, and meanwhile, if the amplitude of the spectral line of the extreme point is the first 30 maximum extreme points, the spectral line is discrete, and the impact fault can be judged.
3. The automatic diagnosis method of rotary mechanical impact fault as claimed in claim 1, wherein the FIR passband filtering method is to sequentially take 1 group of Q point sampling data from P groups of data and supplement K/2 zeros before and after each, and sequentially take K data and filter coefficients from Q+K point data each timeh(k) And multiplying, accumulating and summing, and averaging the summed value, wherein the averaged value is the filtered value.
4. The automatic diagnosis method of rotary machine impact faults as claimed in claim 1, wherein the DFT spectral line calculation method is to obtain a center frequencyf c Spectral line position L c Calculating the initial position of the required analysis frequency spectrum as L c -2 M Analyzing the data position of the spectrum start position in the Q-point FFT resultl ＇ =L c -2 M -Q*INT((L c -2 M ) Q) for use in P sets of Q-point FFT resultsl ＇ The position data is used for obtaining 1 refinement spectrum value according to the DFT principle.
5. The automatic diagnosis method for impact faults of rotary machines according to claim 1, characterized in that the impact spectrum extraction method is divided into a pair of determination L c Will L c The amplitude of each 3 spectral lines is 0, and the spectral line L is calibrated c The corresponding frequency is the zero frequency position of the frequency spectrum coordinate system, and the spectral line L c The spectral lines on the left and right sides are mutually called as spectral lines and are added to obtain an impact spectrum.
6. The automatic diagnosis method for the impact faults of the rotary machine according to claim 1 is characterized in that the impact fault diagnosis method is characterized in that impact frequencies extracted by an impact spectrum extraction method are used for respectively using fault multi-order criteria, fault spectral line discreteness criteria and fault characteristic spectral line energy factor criteria for impact spectrums to effectively eliminate various kinds of interference, so that the accurate identification of the impact faults is realized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110594020.6A CN113567123B (en) | 2021-05-29 | 2021-05-29 | Automatic diagnosis method for impact faults of rotary machinery |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110594020.6A CN113567123B (en) | 2021-05-29 | 2021-05-29 | Automatic diagnosis method for impact faults of rotary machinery |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113567123A CN113567123A (en) | 2021-10-29 |
| CN113567123B true CN113567123B (en) | 2023-09-29 |
Family
ID=78161567
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110594020.6A Active CN113567123B (en) | 2021-05-29 | 2021-05-29 | Automatic diagnosis method for impact faults of rotary machinery |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113567123B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2007263609A (en) * | 2006-03-27 | 2007-10-11 | Nsk Ltd | Apparatus and method for diagnosing failure of mechanical equipment |
| RU2327966C1 (en) * | 2006-10-16 | 2008-06-27 | Государственное образовательное учреждение высшего профессионального образования Воронежский государственный архитектурно-строительный университет (ГОУ ВПО ВГАСУ) | Process of spectral vibration diagnostics of gearing tooth loading in automobiles |
| CN102426102A (en) * | 2011-10-19 | 2012-04-25 | 唐德尧 | Resonance demodulation double isolate frequency spectrum method for detecting crack of gear shaft |
| EP3220120A1 (en) * | 2016-03-17 | 2017-09-20 | ABB Schweiz AG | Method, diagnostic device and system for determining fault conditions in an electrical machine |
| CN108204897A (en) * | 2016-12-16 | 2018-06-26 | 唐智科技湖南发展有限公司 | A kind of bearing parameter correction judgement and multi-parameter diagnose matched method automatically |
| CN111929074A (en) * | 2020-08-19 | 2020-11-13 | 北京经纬恒润科技有限公司 | Vehicle mechanical rotating part fault diagnosis method and device |
| CN112101245A (en) * | 2020-09-18 | 2020-12-18 | 丽水市特种设备检测院 | Short-time Fourier transform mechanical impact feature extraction method based on frequency domain window function |
-
2021
- 2021-05-29 CN CN202110594020.6A patent/CN113567123B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2007263609A (en) * | 2006-03-27 | 2007-10-11 | Nsk Ltd | Apparatus and method for diagnosing failure of mechanical equipment |
| RU2327966C1 (en) * | 2006-10-16 | 2008-06-27 | Государственное образовательное учреждение высшего профессионального образования Воронежский государственный архитектурно-строительный университет (ГОУ ВПО ВГАСУ) | Process of spectral vibration diagnostics of gearing tooth loading in automobiles |
| CN102426102A (en) * | 2011-10-19 | 2012-04-25 | 唐德尧 | Resonance demodulation double isolate frequency spectrum method for detecting crack of gear shaft |
| EP3220120A1 (en) * | 2016-03-17 | 2017-09-20 | ABB Schweiz AG | Method, diagnostic device and system for determining fault conditions in an electrical machine |
| CN108204897A (en) * | 2016-12-16 | 2018-06-26 | 唐智科技湖南发展有限公司 | A kind of bearing parameter correction judgement and multi-parameter diagnose matched method automatically |
| CN111929074A (en) * | 2020-08-19 | 2020-11-13 | 北京经纬恒润科技有限公司 | Vehicle mechanical rotating part fault diagnosis method and device |
| CN112101245A (en) * | 2020-09-18 | 2020-12-18 | 丽水市特种设备检测院 | Short-time Fourier transform mechanical impact feature extraction method based on frequency domain window function |
Non-Patent Citations (2)
| Title |
|---|
| 基于冲击响应谱特征提取的自动机裂纹故障诊断;李海广;潘宏侠;任海锋;兵工学报;37(9);全文 * |
| 短时有效值包络分析在风机轴承故障诊断中的应用;梁双印;潘作为;祖勇海;徐鸿;胡亮;噪声与振动控制;35(4);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113567123A (en) | 2021-10-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP2237003B1 (en) | Method for analysing vibration in rotor blades | |
| CN109883702B (en) | Motor bearing fault diagnosis method based on time-frequency domain statistical characteristics | |
| CN109635334A (en) | Fault Diagnosis of Roller Bearings, system and medium based on particle group optimizing | |
| CN109855874B (en) | Random resonance filter for enhancing detection of weak signals in vibration assisted by sound | |
| CN108151869B (en) | Mechanical vibration characteristic index extraction method, system and device | |
| CN111537881A (en) | Fault diagnosis method, device and equipment for asynchronous motor and readable storage medium | |
| CN109883705A (en) | Motor rolling bearing part spot corrosion method for diagnosing faults and its diagnostic system | |
| CN104359685A (en) | Diesel engine fault identification method | |
| CN112033656A (en) | Mechanical system fault detection method based on broadband spectrum processing | |
| CN114486263A (en) | Noise reduction and demodulation method for vibration signal of rolling bearing of rotary machine | |
| CN113436645A (en) | Electromechanical system fault on-line monitoring acoustic processing method under complex noise environment | |
| CN113567123B (en) | Automatic diagnosis method for impact faults of rotary machinery | |
| JP5436477B2 (en) | Encoder analyzer | |
| CN114486252B (en) | Rolling bearing fault diagnosis method of vector mode maximum envelope | |
| CN116482526A (en) | Analysis system for non-fault phase impedance relay | |
| CN112345247A (en) | Fault diagnosis method and device for rolling bearing | |
| CN113049251A (en) | Bearing fault diagnosis method based on noise | |
| CN114398922A (en) | CNN-BilSTM-based fault diagnosis method for looseness of high-voltage shunt reactor winding | |
| JP2012177653A (en) | Acoustic diagnosis method, program, and device | |
| CN114383718A (en) | High-frequency blade passing frequency extraction method based on vibration signals of external casing of gas turbine | |
| CN113686577A (en) | Bearing fault diagnosis method based on rapid nonlinear sparse spectrum | |
| JP2006125976A (en) | Abnormality diagnosis system of mechanical equipment | |
| CN114090950B (en) | Broadband noise equal-phase compression method and device for stable vibration signals | |
| CN110263711B (en) | Coupling signal impact feature extraction method based on improved spectral kurtosis | |
| CN111079569B (en) | Spindle unbalanced vibration signal processing method based on full-phase FFT method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |