CN111738155A - Bearing clearance fault diagnosis method for reciprocating compressor - Google Patents

Bearing clearance fault diagnosis method for reciprocating compressor Download PDF

Info

Publication number
CN111738155A
CN111738155A CN202010581666.6A CN202010581666A CN111738155A CN 111738155 A CN111738155 A CN 111738155A CN 202010581666 A CN202010581666 A CN 202010581666A CN 111738155 A CN111738155 A CN 111738155A
Authority
CN
China
Prior art keywords
signal
scale
empirical
mode
bearing
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.)
Pending
Application number
CN202010581666.6A
Other languages
Chinese (zh)
Inventor
王金东
李彦阳
赵海洋
张隆宇
陈新
于德龙
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Northeast Petroleum University
Original Assignee
Northeast Petroleum University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Northeast Petroleum University filed Critical Northeast Petroleum University
Priority to CN202010581666.6A priority Critical patent/CN111738155A/en
Publication of CN111738155A publication Critical patent/CN111738155A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching
    • G06F2218/16—Classification; Matching by matching signal segments
    • 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
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
    • G06F17/148—Wavelet transforms
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Mathematical Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Computational Mathematics (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Algebra (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Signal Processing (AREA)
  • Acoustics & Sound (AREA)
  • Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)

Abstract

The invention relates to a reciprocating compressor bearing clearance fault diagnosis method, which comprises the steps of improving a frequency spectrum separation method of empirical wavelet transform, taking variance as an evaluation parameter, and preferably selecting a scale transformation parameter; secondly, dividing the Fourier spectrum of the vibration signal by applying the optimized scale transformation parameters to obtain an empirical wavelet mode; thirdly, calculating a Pearson correlation coefficient and a modal kurtosis between an empirical wavelet mode and an original signal, and screening the empirical wavelet mode; determining a self-adaptive energy threshold for dividing the signal state, and judging the state of the signal according to the self-adaptive energy threshold and segmenting the signal; constructing structural elements according to different states of the signals, classifying the segmented energy, and constructing the structural elements according to different states for filtering; step six: and finally, carrying out quantitative analysis by using the form spectrum entropy to finish the classification and identification of the fault. The method for obtaining the position of the local minimum value by utilizing the combined action of the multi-scale information is more accurate and has strong anti-interference capability.

Description

Bearing clearance fault diagnosis method for reciprocating compressor
Technical Field
The invention relates to the technical field of reciprocating mechanical fault diagnosis, in particular to a reciprocating compressor bearing clearance fault diagnosis method based on self-adaptive empirical wavelet transform and state morphology filtering.
Background
The petroleum and chemical industry is an important foundation and pillar industry of national economy, and reciprocating compressors are important equipment widely used in the field. The crankshaft, connecting rod and crosshead of reciprocating compressor are the important components of transmission mechanism, and their functions are to convert the rotary motion of driving motor into reciprocating linear motion of piston and to undertake the power transmission function of whole machine set. The crosshead bearing and the crankshaft bearing are impacted in each reversing process of the reciprocating compressor, and the bearings are always in a friction and collision state due to the fact that the bearings bear large alternating load and friction abrasion, and the bearing gap is inevitably enlarged after long-term operation, so that the vibration amplitude is increased, the abrasion is aggravated after the vibration amplitude is increased, and the serious circulation can cause breakage accidents of key parts such as a crankshaft, a connecting rod, a piston rod and the like if the serious circulation is not found in time. If the bearing has defects of design, manufacture and the like, and problems of poor lubrication, excessive impact force and the like, the bearing is abraded or failed. Therefore, the method has important engineering application value for researching the reciprocating compressor bearing clearance fault diagnosis method.
The vibration signal of the reciprocating compressor belongs to a non-stationary and non-linear signal. Common methods currently used to process non-stationary and non-linear signals are Empirical Mode Decomposition (EMD) and Empirical Wavelet Transform (EWT). The EMD has no strict mathematical proof when processing nonlinear non-stationary signals, and has the defects of over-enveloping, under-enveloping and large calculation amount. The EWT is a self-adaptive wavelet analysis method, but has the problems that the separation modes are too many, and the obtained single-component mode cannot be subjected to Hilbert transformation.
Wavelet threshold denoising has been developed as the most commonly used filtering method through research and improvement of numerous scholars. However, the threshold selection of wavelet filtering is relatively different for different types of signals, and for nonlinear and non-stationary signals, only one wavelet basis function is used for processing, so that the characteristics of the signals cannot be well reserved.
Disclosure of Invention
The invention aims to provide a reciprocating compressor bearing clearance fault diagnosis method which is used for solving the problem of reciprocating compressor bearing clearance fault diagnosis.
The technical scheme adopted by the invention for solving the technical problems is as follows: the reciprocating compressor bearing clearance fault diagnosis method comprises the following steps:
the method comprises the following steps: based on a scale space theory, improving a spectrum separation method of empirical wavelet transform, taking a square difference as an evaluation parameter, judging whether a spectrum boundary is stable, and carrying out comparative analysis on a plurality of groups of reciprocating compressor bearing gap fault vibration signal experimental data to preferably select scale transformation parameters;
step two: dividing the Fourier spectrum of the vibration signal by applying the optimized scale transformation parameters to obtain an empirical wavelet mode;
1) for a reciprocating compressor, because a crank case and a crosshead are sensitive to bearing clearance faults, 4 sensors are respectively arranged at a first-stage connecting rod crosshead, a second-stage connecting rod crosshead, a first-stage connecting rod crank case and a second-stage connecting rod crank case to acquire bearing clearance fault vibration signals, the sampling frequency is 50kHz, and the sampling time is 4 s; selecting a bearing clearance fault vibration signal of 6000 points in a whole period for analysis, wherein the analysis comprises four fault states: (a) the first-stage small-head bearing has large clearance, (b) the first-stage big-head bearing has large clearance, (c) the second-stage small-head bearing has large clearance, (d) the second-stage big-head bearing has large clearance;
2) performing empirical wavelet transform based on a scale space theory, and performing adaptive decomposition on the vibration signal by using an optimal scale transform parameter to obtain a frequency spectrum separation boundary;
3) constructing an empirical scale function and an empirical wavelet function by using the Meyer wavelet to obtain an orthogonal wavelet filter; performing empirical wavelet transform on the frequency spectrum to obtain a group of AM-FM components of single components or approximate single components;
4) hilbert transformation is carried out on the group of AM-FM components obtained in the step 3), time-frequency analysis is carried out on signals through Hilbert spectrums, frequency spectrum separation is firstly carried out, and then normalization processing is carried out on empirical modes obtained through decomposition;
step three: calculating a Pearson correlation coefficient and a modal kurtosis between an empirical wavelet mode and an original signal, and screening the empirical wavelet mode;
step four: determining a self-adaptive energy threshold for dividing the signal state, and judging the state of the signal according to the self-adaptive energy threshold and segmenting the signal;
the sampling frequency is 120 points when being 50KHz, and 6000 points in the whole period are divided into 200 segments for analysis; respectively calculating the sum of absolute values of the energy of each section of signal to serve as an energy threshold initial judgment index; when the energy of a certain section is not much more than that of a stationary section, the K-means algorithm is used as an energy threshold value fine judgment index; the segmented energy is a bearing gap vibration state signal when the energy is higher than the energy threshold, and a bearing gap stable state signal when the energy is lower than the energy threshold;
step five: according to different states of the signal, structural elements are constructed, the segmented energy is classified, and the structural elements are constructed according to different states for filtering, and the method specifically comprises the following steps:
1) and constructing structural elements according to different states for filtering: the flat structural element is used for the vibration state of the bearing gap signal, and the shape characteristic of an impact signal is kept to the maximum extent while filtering is carried out; the triangular structural elements are used for the stable state of a bearing gap signal, and pulse signals with large interference on the signal are filtered;
2) respectively calculating the average of the absolute value of the amplitude of each section of signal to be used as the height of the structural element of the section, and selecting the length of the structural element by combining the state filtering segmentation condition;
3) in order to maintain the morphological characteristics of the signal, morphological filtering is carried out by using an open-plus-closed-mean filter;
step six: and finally, carrying out quantitative analysis by using the form spectrum entropy, and completing the classification and identification of the fault by drawing four state form spectrum entropy graphs of the bearing clearance fault.
The specific method of the step one in the scheme comprises the following steps: decomposing the collected reciprocating compressor bearing clearance fault vibration acceleration signal, analyzing 1-10000Hz, wherein the sampling frequency of an experimental signal is 25000Hz, and the frequency of more than 10000Hz is high-frequency noise; judging whether the spectrum boundary is stable or not by taking the variance as an evaluation parameter, calculating the modal boundaries of which the scale transformation parameters are respectively 0.5, 0.75, 1, 1.5 and 2, and reserving the first five separated modal boundaries; through comparative analysis, when the scale transformation parameter is 1, the variance between the boundaries is minimum; the comparative analysis of the experimental data of the bearing clearance fault vibration signals of the 30 groups of reciprocating compressors shows that when the scale transformation parameter is 1, the values of all modal boundaries are basically consistent, and the frequency spectrum separation boundary has stability.
The specific steps of step 2) in the second step of the scheme are as follows:
2.1) applying a scale-space approach, firstly using a gaussian kernel function to perform scale transformation on the signal:
Figure BDA0002553409430000031
wherein
Figure BDA0002553409430000032
The finite impulse response filter needs to be obtained using a shortened filter, so the value of n is controlled:
Figure BDA0002553409430000033
wherein
Figure BDA0002553409430000034
C6 to ensure that the approximation error is less than 10-9；
f, (x) obtaining a frequency domain function after Fourier transform of the vibration signal;
g (n: t) is a Gaussian kernel function;
n, obtaining a finite-length impulse response filter by controlling the value of M according to the position parameter in the scale transformation;
t is a scale parameter in the scale transformation, which consists of an initial scale parameter and a scale transformation parameter;
2.2) obtaining different scale spaces after scale transformation, recording the minimum value position of the signal after each space transformation, and connecting the minimum values in a scale space plane to obtain a scale space minimum value curve; and selecting a threshold value through a K-means classification algorithm, reserving a scale space minimum value curve larger than the threshold value, and determining the position of the curve as a spectrum separation boundary.
The step 3) in the second step in the scheme comprises the following specific steps:
3.1) Meyer wavelet construction empirical scale function is:
Figure BDA0002553409430000041
3.2) empirical wavelet function as:
Figure BDA0002553409430000042
wherein β (x) ═ x4(35-84x+70x2-20x3)
τn＝γωn
Figure BDA0002553409430000043
ωnThe spectrum separation boundary determined in 2.2), gamma and tau are only process parameters for constructing an empirical function, and have no specific meaning;
3.3) determining detail coefficients and approximation coefficients respectively by using an inner product method:
Figure BDA0002553409430000051
Figure BDA0002553409430000052
3.4) empirical mode function can be expressed as:
Figure BDA0002553409430000053
Figure BDA0002553409430000054
Ψ is the empirical wavelet function calculated in 3.2), φ is the empirical scaling function calculated in 3.1), denotes the convolution operation.
The concrete method of the third step in the scheme is as follows:
1) the pearson correlation coefficient describes how closely two variables are linked, and is formulated as follows:
Figure BDA0002553409430000055
wherein
Figure BDA0002553409430000056
Is the mean value of the original signal samples;
Figure BDA0002553409430000057
Figure BDA0002553409430000058
to a calculated modal signal sample mean;
Figure BDA0002553409430000059
Figure BDA00025534094300000510
in order to be a sample of the original signal,
Figure BDA00025534094300000511
for the calculated modal signal sample, T is the signal sequence length;
2) optimizing the mode by calculating the correlation degree of the mode and the original signal, eliminating the distorted mode, and keeping the empirical mode with the Pearson correlation coefficient larger than 0.4;
3) the kurtosis index is particularly sensitive to characterizing impacts, and a mode preference criterion is adopted: the mode with the highest kurtosis is further selected from all modes with Pearson correlation coefficients larger than 0.4.
The invention has the following beneficial effects:
1. the invention provides a reciprocating compressor bearing clearance fault diagnosis method based on self-adaptive empirical wavelet transform and state morphology filtering, which is based on a scale space theory, optimizes scale parameters and improves a frequency spectrum separation method of empirical wavelet transform. The method for obtaining the position of the local minimum value by utilizing the combined action of the multi-scale information is more accurate than directly finding the local minimum value on the original frequency spectrum, and has strong anti-jamming capability. Aiming at the defects of wavelet filtering, a filtering method of state morphology is adopted, and the signals are matched by continuously moving structural elements in the signals, so that the purposes of extracting the signals, keeping details and suppressing noise are achieved. The method can effectively process the characteristics of the signals such as eliminating positive and negative impacts, reducing noise and the like, has simpler and more convenient algorithm and higher calculation speed, and is easy to realize hardware.
2. Aiming at the characteristics of complex nonlinearity, non-stability and the like of a reciprocating compressor vibration signal, firstly, a scale space curve is constructed by using self-adaptive empirical wavelet transform to divide a Fourier spectrum, and an appropriate orthogonal wavelet filter bank is constructed to extract an AM-FM component with a tightly-supported Fourier spectrum; aiming at the problem that the empirical wavelet transform has too many separation modes and the obtained single-component mode can not be subjected to Hibert transform, selecting the mode with the most obvious vibration impact characteristic by calculating the Pearson correlation coefficient and the modal kurtosis between the empirical wavelet mode and an original signal, thereby realizing mode optimization; aiming at the problem that a reciprocating mechanism transmission device has a large amount of impact and cannot adapt to impact waveforms and non-impact waveforms simultaneously if a single structural element is used for filtering, the bearing clearance fault states are classified by a state morphology method, and the structural elements are adaptively constructed according to different states of signals, so that the states are filtered and fault features are extracted; and finally, the fault type is more accurately diagnosed through the form spectrum entropy.
3. The invention carries out modal division on improved empirical wavelet transform, and carries out fault identification by using the morphological entropy after state morphological filtering, and aims to solve the problem of fault diagnosis of the bearing clearance of the reciprocating compressor.
Drawings
FIG. 1 is a flow chart of the diagnostic method of the present invention.
FIG. 2 is a time domain waveform of vibration signals in four states of a reciprocating compressor bearing clearance fault, wherein (a) a first-stage small-head bearing clearance is large; (b) the first-level big head bearing has large clearance; (c) the second-stage small-head bearing has large clearance; (d) the second-level big head bearing has large clearance.
FIG. 3 is a reciprocating compressor bearing clearance fault frequency spectrum separation boundary diagram wherein (a) the primary big head bearing clearance is large; (b) the second-stage small-head bearing has large clearance.
FIG. 4 is a mode screened for reciprocating compressor bearing clearance failure wherein (a) the first stage big head bearing clearance is large; (b) the second-stage small-head bearing has large clearance.
FIG. 5 is a modal shape after fault filtering of reciprocating compressor bearing clearances, wherein (a) the first stage big end bearing clearances are large; (b) the second-stage small-head bearing has large clearance.
FIG. 6 is a morphological entropy of four states of a reciprocating compressor bearing clearance failure.
Detailed Description
The invention is further described with reference to the accompanying drawings in which:
as shown in fig. 1, the reciprocating compressor bearing clearance fault diagnosis method comprises the following steps:
the method comprises the following steps: based on the scale space theory, the spectrum separation method of the empirical wavelet transform is improved, and scale change parameters are optimized;
collecting vibration acceleration signals of bearing clearance faults of the reciprocating compressor, and performing wavelet decomposition with a sampling frequency of 25000 Hz. High frequency noise is above 10000Hz, so only 1-10000Hz is analyzed. The spectrum separation method of the empirical wavelet transform is improved, and scale change parameters are optimized. The method comprises the following specific steps:
and judging whether the spectrum boundary is stable or not by taking the variance as an evaluation parameter. The variances of the modal separation boundaries for the scaling parameters 0.5, 0.75, 1, 1.5, 2, respectively, were calculated, as shown in table 1, with the first five separated modal boundaries retained. As can be seen from table 1, when the scaling parameter is 1, the variance between the boundaries is minimal.
TABLE 1 variance of modal separation boundaries for parameters of each scale
Boundary 1 Boundary 2 Boundary 3 Boundary 4 Boundary 5
0.5 408 1067 1215 2039 1938
0.75 91 447 722 1445 1808
1 91 433 339 1089 1493
1.5 91 1552 2055 1850 2168
2 2247 2616 3183 3601 5409
In order to macroscopically compare the difference of the boundaries between the states, 30 groups of reciprocating compressor bearing clearance fault vibration signal experimental data are contrasted and analyzed, and when the scale transformation parameter is 1, the modal separation boundary is shown in the table 2. Comparing each column of data, and finding that the values of the modal boundaries are basically consistent. It is thus demonstrated that the spectral separation boundary has stability when the scaling parameter is 1;
TABLE 2 bearing Clearance Fault vibration Signal Modal separation boundary with Scale parameter 1
Figure BDA0002553409430000071
Figure BDA0002553409430000081
Step two: and dividing the Fourier spectrum of the signal by applying the optimized parameters to obtain an empirical wavelet mode.
The method comprises the following specific steps:
the double-acting reciprocating compressor of 2D12-70 type is used as a research object, because a crank case and a crosshead are sensitive to bearing clearance faults, 4 sensors are respectively arranged at a first-stage connecting rod crosshead, a second-stage connecting rod crosshead, a first-stage connecting rod crank case and a second-stage connecting rod crank case to collect vibration signals, the sampling frequency is 50kHz, the sampling time is 4s, and the vibration signals of the bearing clearance faults of a whole period (6000 points) are selected for analysis, and the double-acting reciprocating compressor comprises four fault states: (a) the first-stage small-head bearing has large clearance, (b) the first-stage big-head bearing has large clearance, (c) the second-stage small-head bearing has large clearance, (d) the second-stage big-head bearing has large clearance. As shown in fig. 2.
And performing empirical wavelet transform based on a scale space theory, and performing adaptive decomposition on the vibration signal by using an optimal scale transform parameter 1 to obtain a frequency spectrum separation boundary. The calculation steps are as follows:
applying a scale space method, firstly using a Gaussian kernel function to carry out scale transformation on a signal:
Figure BDA0002553409430000082
wherein
Figure BDA0002553409430000083
In practical applications, we need to use a shortened filter to obtain a finite impulse response filter, so we control the value of n:
Figure BDA0002553409430000091
wherein
Figure BDA0002553409430000092
According to the analysis of experimental data, C is 6 to ensure that the approximation error is less than 10-9. And obtaining different scale spaces after scale transformation, recording the minimum value positions of the signals after each space transformation, and connecting the minimum values in a scale space plane to obtain a scale space minimum value curve. And selecting a threshold value through a K-means classification algorithm, reserving a scale space minimum value curve larger than the threshold value, and determining the position of the curve as a spectrum separation boundary.
And constructing an empirical scale function and an empirical wavelet function by using the Meyer wavelet to obtain the orthogonal wavelet filter. And performing empirical wavelet transform on the frequency spectrum to obtain a group of single-component or approximate single-component AM-FM components.
The empirical scale function of the Meyer wavelet construction is as follows:
Figure BDA0002553409430000093
the empirical wavelet function is:
Figure BDA0002553409430000094
wherein β (x) ═ x4(35-84x+70x2-20x3)
τn＝γωn
Figure BDA0002553409430000095
The detail coefficients and the approximation coefficients are determined separately using an inner product method:
Figure BDA0002553409430000101
Figure BDA0002553409430000102
the empirical mode function may be expressed as:
Figure BDA0002553409430000103
Figure BDA0002553409430000104
hilbert transform is performed on the group of AM-FM components, and signals are subjected to time-frequency analysis through Hilbert spectrum: firstly, performing frequency spectrum separation, wherein a boundary graph of the frequency spectrum separation of the first-stage large-head bearing with large clearance and the second-stage small-head bearing with large clearance is shown in figure 3, and then performing normalization processing on the empirical mode obtained by decomposition.
Step three: calculating the Peak of Pearson correlation coefficient and mode between empirical wavelet mode and original signal, and screening empirical wavelet mode, which comprises the following steps:
the pearson correlation coefficient describes how closely two variables are linked, and is formulated as follows:
Figure BDA0002553409430000105
Figure BDA0002553409430000106
Figure BDA0002553409430000107
and (4) optimizing the mode by calculating the correlation degree of the mode and the original signal, and eliminating the distorted mode. The empirical mode with a pearson correlation coefficient greater than 0.4 is retained.
The kurtosis index is particularly sensitive to impact characteristics in signals, and a mode with the highest kurtosis is further selected from all modes with Pearson correlation coefficients larger than 0.4 by adopting a mode preference criterion. And performing mode screening according to the indexes, and reserving modes with more fault information, wherein the screened modes with the large first-stage large head bearing gap and the large second-stage small head bearing gap are shown in FIG. 4.
Step four: determining a self-adaptive energy threshold for dividing the signal state, judging the state of the signal according to the self-adaptive energy threshold, and segmenting, wherein the method comprises the following specific steps:
the sampling frequency is about 120 points when being 50 KHz. 6000 points of the whole cycle were divided into 200 segments for analysis.
And determining the self-adaptive energy threshold for dividing the signal state by combining multiple groups of experimental data and the characteristics of the signal: and respectively calculating the sum of the absolute values of the energy of each section of signal to serve as an energy threshold initial judgment index. And when the energy of a certain section is not much more than that of the stationary section, the K-means algorithm is used as a fine judgment energy threshold index. And the segmented energy is higher than the energy threshold value and is a bearing clearance fault vibration state signal, and the segmented energy is lower than the energy threshold value and is a bearing clearance fault stable state signal.
Step five: constructing proper structural elements according to different states for filtering: the flat structural element is used for the vibration state of the bearing clearance fault signal, and the shape characteristic of the impact signal is kept to the maximum extent while filtering is carried out. The triangular structural elements are used for the stable state of the bearing clearance fault signal, and pulse signals with large interference to the signals are filtered.
And respectively calculating the average of the absolute value of the amplitude of each section of signal to be used as the height of the structural element of the section, and selecting the length of the structural element to be 3 by combining the state filtering segmentation condition. To preserve the morphological characteristics of the signal, morphological filtering was performed by multiple experimental comparisons using an open + closed mean filter. The empirical mode components after filtering are shown in fig. 5 when the first-stage large-head bearing clearance is large and the second-stage small-head bearing clearance is large.
Step six: and finally, carrying out quantitative analysis by using the morphological spectrum entropy. Form spectrum entropies of four states of the bearing clearance fault are respectively calculated, as shown in fig. 6, the four states are clearly distinguished: 30 sets of data for each state, for a total of 120 sets of data. Wherein, 1-30 groups are in a state of large clearance of the first-stage small-head bearing, 31-60 groups are in a state of large clearance of the first-stage big-head bearing, 61-90 groups are in a state of large clearance of the second-stage small-head bearing, and 91-120 groups are in a state of large clearance of the second-stage big-head bearing. The form spectrum entropy range of the first-stage small-head bearing in the large clearance state is centered on 0.754 and floats up and down by about 0.02; the form spectrum entropy range of the first-stage big head bearing in a large clearance state is centered at 0.793 and fluctuates up and down by about 0.03; the form spectrum entropy range with large second-stage small-head bearing clearance is centered at 0.707 and floats up and down by about 0.02; the form spectrum entropy range with large second-level big head bearing clearance is centered at 0.66 and fluctuates up and down by about 0.025.
The identification rate of the method for the large clearance state of the first-stage small-head bearing is 93.33%, the identification rate of the large clearance state of the first-stage large-head bearing is 96.66%, the identification rate of the large clearance state of the second-stage small-head bearing is 96.66%, the identification rate of the large clearance state of the second-stage large-head bearing is 90%, and the effectiveness of the method for accurately identifying the clearance fault of the bearing is verified.
The above description is only for the best mode of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the appended claims.

Claims (5)

1. A reciprocating compressor bearing clearance fault diagnosis method is characterized by comprising the following steps:
the method comprises the following steps: based on a scale space theory, improving a spectrum separation method of empirical wavelet transform, taking a square difference as an evaluation parameter, judging whether a spectrum boundary is stable, and carrying out comparative analysis on a plurality of groups of reciprocating compressor bearing gap fault vibration signal experimental data to preferably select scale transformation parameters;
step two: dividing the Fourier spectrum of the vibration signal by applying the optimized scale transformation parameters to obtain an empirical wavelet mode;
1) for a reciprocating compressor, because a crank case and a crosshead are sensitive to bearing clearance faults, 4 sensors are respectively arranged at a first-stage connecting rod crosshead, a second-stage connecting rod crosshead, a first-stage connecting rod crank case and a second-stage connecting rod crank case to acquire bearing clearance fault vibration signals, the sampling frequency is 50kHz, and the sampling time is 4 s; selecting a bearing clearance fault vibration signal of 6000 points in a whole period for analysis, wherein the analysis comprises four fault states: (a) the first-stage small-head bearing has large clearance, (b) the first-stage big-head bearing has large clearance, (c) the second-stage small-head bearing has large clearance, (d) the second-stage big-head bearing has large clearance;
2) performing empirical wavelet transform based on a scale space theory, and performing adaptive decomposition on the vibration signal by using an optimal scale transform parameter to obtain a frequency spectrum separation boundary;
3) constructing an empirical scale function and an empirical wavelet function by using the Meyer wavelet to obtain an orthogonal wavelet filter; performing empirical wavelet transform on the frequency spectrum to obtain a group of AM-FM components of single components or approximate single components;
4) hilbert transformation is carried out on the group of AM-FM components obtained in the step 3), time-frequency analysis is carried out on signals through Hilbert spectrums, frequency spectrum separation is firstly carried out, and then normalization processing is carried out on empirical modes obtained through decomposition;
step three: calculating a Pearson correlation coefficient and a modal kurtosis between an empirical wavelet mode and an original signal, and screening the empirical wavelet mode;
step four: determining a self-adaptive energy threshold for dividing the signal state, and judging the state of the signal according to the self-adaptive energy threshold and segmenting the signal;
the sampling frequency is 120 points when being 50KHz, and 6000 points in the whole period are divided into 200 segments for analysis; respectively calculating the sum of absolute values of the energy of each section of signal to serve as an energy threshold initial judgment index; when the energy of a certain section is not much more than that of a stationary section, the K-means algorithm is used as an energy threshold value fine judgment index; the segmented energy is a bearing gap vibration state signal when the energy is higher than the energy threshold, and a bearing gap stable state signal when the energy is lower than the energy threshold;
step five: according to different states of the signal, structural elements are constructed, the segmented energy is classified, and the structural elements are constructed according to different states for filtering, and the method specifically comprises the following steps:
1) and constructing structural elements according to different states for filtering: the flat structural element is used for the vibration state of the bearing gap signal, and the shape characteristic of an impact signal is kept to the maximum extent while filtering is carried out; the triangular structural elements are used for the stable state of a bearing gap signal, and pulse signals with large interference on the signal are filtered;
2) respectively calculating the average of the absolute value of the amplitude of each section of signal to be used as the height of the structural element of the section, and selecting the length of the structural element by combining the state filtering segmentation condition;
3) in order to maintain the morphological characteristics of the signal, morphological filtering is carried out by using an open-plus-closed-mean filter;
step six: and finally, carrying out quantitative analysis by using the form spectrum entropy, and completing the classification and identification of the fault by drawing four state form spectrum entropy graphs of the bearing clearance fault.
2. The reciprocating compressor bearing clearance fault diagnosis method of claim 1, wherein: the specific method of the first step is as follows: decomposing the collected reciprocating compressor bearing clearance fault vibration acceleration signal, analyzing 1-10000Hz, wherein the sampling frequency of an experimental signal is 25000Hz, and the frequency of more than 10000Hz is high-frequency noise; judging whether the spectrum boundary is stable or not by taking the variance as an evaluation parameter, calculating the modal boundaries of which the scale transformation parameters are respectively 0.5, 0.75, 1, 1.5 and 2, and reserving the first five separated modal boundaries; through comparative analysis, when the scale transformation parameter is 1, the variance between the boundaries is minimum; the comparative analysis of the experimental data of the bearing clearance fault vibration signals of the 30 groups of reciprocating compressors shows that when the scale transformation parameter is 1, the values of all modal boundaries are basically consistent, and the frequency spectrum separation boundary has stability.
3. The reciprocating compressor bearing clearance fault diagnosis method of claim 2, wherein: the specific steps of step 2) in the second step:
2.1) applying a scale-space approach, firstly using a gaussian kernel function to perform scale transformation on the signal:
Figure FDA0002553409420000021
wherein
Figure FDA0002553409420000022
The finite impulse response filter needs to be obtained using a shortened filter, so the value of n is controlled:
Figure FDA0002553409420000023
wherein
Figure FDA0002553409420000024
C6 to ensure that the approximation error is less than 10-9；
f, (x) obtaining a frequency domain function after Fourier transform of the vibration signal;
g (n: t) is a Gaussian kernel function;
n, obtaining a finite-length impulse response filter by controlling the value of M according to the position parameter in the scale transformation;
t is a scale parameter in the scale transformation, which consists of an initial scale parameter and a scale transformation parameter;
2.2) obtaining different scale spaces after scale transformation, recording the minimum value position of the signal after each space transformation, and connecting the minimum values in a scale space plane to obtain a scale space minimum value curve; and selecting a threshold value through a K-means classification algorithm, reserving a scale space minimum value curve larger than the threshold value, and determining the position of the curve as a spectrum separation boundary.
4. The reciprocating compressor bearing clearance fault diagnosis method of claim 3, wherein: the specific step of step 3) in the second step:
3.1) Meyer wavelet construction empirical scale function is:
Figure FDA0002553409420000031
3.2) empirical wavelet function as:
Figure FDA0002553409420000032
wherein β (x) ═ x4(35-84x+70x2-20x3)
τn＝γωn
Figure FDA0002553409420000033
ωnThe spectrum separation boundary determined in 2.2), gamma and tau are only process parameters for constructing an empirical function, and have no specific meaning;
3.3) determining detail coefficients and approximation coefficients respectively by using an inner product method:
Figure FDA0002553409420000041
Figure FDA0002553409420000042
3.4) empirical mode function can be expressed as:
Figure FDA0002553409420000043
Figure FDA0002553409420000044
Ψ is the empirical wavelet function calculated in 3.2), φ is the empirical scaling function calculated in 3.1), denotes the convolution operation.
5. The reciprocating compressor bearing clearance fault diagnosis method of claim 4, wherein: the concrete method of the third step:
1) the pearson correlation coefficient describes how closely two variables are linked, and is formulated as follows:
Figure FDA0002553409420000045
wherein
Figure FDA0002553409420000046
Mean of raw signal samples:
Figure FDA0002553409420000047
Figure FDA0002553409420000048
to calculate the modal signal sample mean:
Figure FDA0002553409420000049
Figure FDA00025534094200000410
in order to be a sample of the original signal,
Figure FDA00025534094200000411
for the calculated modal signal sample, T is the signal sequence length;
2) optimizing the mode by calculating the correlation degree of the mode and the original signal, eliminating the distorted mode, and keeping the empirical mode with the Pearson correlation coefficient larger than 0.4;
3) the kurtosis index is particularly sensitive to characterizing impacts, and a mode preference criterion is adopted: the mode with the highest kurtosis is further selected from all modes with Pearson correlation coefficients larger than 0.4.
CN202010581666.6A 2020-06-23 2020-06-23 Bearing clearance fault diagnosis method for reciprocating compressor Pending CN111738155A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010581666.6A CN111738155A (en) 2020-06-23 2020-06-23 Bearing clearance fault diagnosis method for reciprocating compressor

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010581666.6A CN111738155A (en) 2020-06-23 2020-06-23 Bearing clearance fault diagnosis method for reciprocating compressor

Publications (1)

Publication Number Publication Date
CN111738155A true CN111738155A (en) 2020-10-02

Family

ID=72650677

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010581666.6A Pending CN111738155A (en) 2020-06-23 2020-06-23 Bearing clearance fault diagnosis method for reciprocating compressor

Country Status (1)

Country Link
CN (1) CN111738155A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113049252A (en) * 2021-03-25 2021-06-29 成都天佑路航轨道交通科技有限公司 Fault detection method for train bearing box
CN113740066A (en) * 2021-11-08 2021-12-03 中国空气动力研究与发展中心设备设计与测试技术研究所 Early fault detection method for compressor bearing
CN114757060A (en) * 2022-06-15 2022-07-15 石家庄铁道大学 Bearing fault diagnosis method based on morphological function construction scale space

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103439110A (en) * 2013-07-31 2013-12-11 哈尔滨工程大学 Early-stage weak fault diagnosis method of rolling bearing
CN109707615A (en) * 2019-02-26 2019-05-03 东北石油大学 Reciprocating compressor method for diagnosing faults based on fine multi-fractal
CN111141520A (en) * 2020-02-24 2020-05-12 江南大学 Rolling bearing fault diagnosis method based on improved experience wavelet transform
CN111259992A (en) * 2020-03-03 2020-06-09 东北石油大学 Reciprocating compressor air valve fault diagnosis method based on time-varying singular spectrum

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103439110A (en) * 2013-07-31 2013-12-11 哈尔滨工程大学 Early-stage weak fault diagnosis method of rolling bearing
CN109707615A (en) * 2019-02-26 2019-05-03 东北石油大学 Reciprocating compressor method for diagnosing faults based on fine multi-fractal
CN111141520A (en) * 2020-02-24 2020-05-12 江南大学 Rolling bearing fault diagnosis method based on improved experience wavelet transform
CN111259992A (en) * 2020-03-03 2020-06-09 东北石油大学 Reciprocating compressor air valve fault diagnosis method based on time-varying singular spectrum

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
张隆宇: "《经验小波振动特征提取及在往复压缩机故障诊断中的应用》", 《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》 *
张鲁洋 等: "《基于两类特征与AFSA优化SVM的滚动轴承故障诊断研究》", 《机械强度》 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113049252A (en) * 2021-03-25 2021-06-29 成都天佑路航轨道交通科技有限公司 Fault detection method for train bearing box
CN113049252B (en) * 2021-03-25 2023-04-14 成都天佑路航轨道交通科技有限公司 Fault detection method for train bearing box
CN113740066A (en) * 2021-11-08 2021-12-03 中国空气动力研究与发展中心设备设计与测试技术研究所 Early fault detection method for compressor bearing
CN113740066B (en) * 2021-11-08 2022-02-08 中国空气动力研究与发展中心设备设计与测试技术研究所 Early fault detection method for compressor bearing
CN114757060A (en) * 2022-06-15 2022-07-15 石家庄铁道大学 Bearing fault diagnosis method based on morphological function construction scale space

Similar Documents

Publication Publication Date Title
CN111738155A (en) Bearing clearance fault diagnosis method for reciprocating compressor
Li et al. A weighted multi-scale morphological gradient filter for rolling element bearing fault detection
Lei et al. A review on empirical mode decomposition in fault diagnosis of rotating machinery
Gao et al. Non-stationary signal processing for bearing health monitoring
Hao et al. Application of support vector machine based on pattern spectrum entropy in fault diagnostics of rolling element bearings
CN104634571A (en) Fault diagnosis method for rolling bearing based on LCD-MF (Local Characteristic Scale Decomposition )-(Multifractal)
Guo et al. An enhanced modulation signal bispectrum analysis for bearing fault detection based on non-Gaussian noise suppression
CN113639999A (en) Fault diagnosis method, system, medium and equipment for rolling bearing under variable rotating speed
Jiang et al. A novel method for self-adaptive feature extraction using scaling crossover characteristics of signals and combining with LS-SVM for multi-fault diagnosis of gearbox
Bastami et al. Estimating the size of naturally generated defects in the outer ring and roller of a tapered roller bearing based on autoregressive model combined with envelope analysis and discrete wavelet transform
Chen et al. A data-driven threshold for wavelet sliding window denoising in mechanical fault detection
Chen et al. A signal based triangular structuring element for mathematical morphological analysis and its application in rolling element bearing fault diagnosis
CN114354194B (en) Rolling bearing fault diagnosis method based on full convolution self-encoder and optimized support vector machine
Ambika et al. Vibration signal based condition monitoring of mechanical equipment with scattering transform
Wang et al. Diesel engine gearbox fault diagnosis based on multi-features extracted from vibration signals
Gong et al. Fault detection for rolling element bearing based on repeated single-scale morphology and simplified sensitive factor algorithm
Fang et al. Reciprocating compressors intelligent fault diagnosis under multiple operating conditions based on adaptive variable scale morphological filter
CN112419304B (en) Multi-stage target detection method and device for one-dimensional data
Yan et al. A chaotic feature extraction based on SMMF and CMMFD for early fault diagnosis of rolling bearing
CN112747925B (en) Rolling bearing fault diagnosis method based on composite morphological filtering
CN114705427A (en) Rolling bearing intelligent fault diagnosis method under strong noise environment
Nayana et al. Effective time domain features for identification of bearing fault using LDA and NB classifiers
Zhao et al. An early fault diagnosis method of rolling element bearings based on med, dfa, and improved knn
Yuan et al. Gearbox fault diagnosis of rolling mills using multiwavelet sliding window neighboring coefficient denoising and optimal blind deconvolution
Shang et al. Denoising Fused Wavelets Net for Aeroengine Bevel Gear Fault Diagnosis

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20201002