CN114091523A - Method for diagnosing gray fault of key rotating part of signal frequency domain characteristic driven vehicle - Google Patents

Method for diagnosing gray fault of key rotating part of signal frequency domain characteristic driven vehicle Download PDF

Info

Publication number
CN114091523A
CN114091523A CN202111191225.6A CN202111191225A CN114091523A CN 114091523 A CN114091523 A CN 114091523A CN 202111191225 A CN202111191225 A CN 202111191225A CN 114091523 A CN114091523 A CN 114091523A
Authority
CN
China
Prior art keywords
frequency domain
gray
band
frequency
sub
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
CN202111191225.6A
Other languages
Chinese (zh)
Inventor
石娟娟
苏舟
郑传玺
胡丽敏
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.)
Jiangsu Ktk Locomotive & Rolling Stock Co ltd
Original Assignee
Jiangsu Ktk Locomotive & Rolling Stock Co ltd
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 Jiangsu Ktk Locomotive & Rolling Stock Co ltd filed Critical Jiangsu Ktk Locomotive & Rolling Stock Co ltd
Priority to CN202111191225.6A priority Critical patent/CN114091523A/en
Publication of CN114091523A publication Critical patent/CN114091523A/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
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
    • G01H17/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves, not provided for in the preceding groups
    • 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
    • G01M13/02—Gearings; Transmission mechanisms
    • G01M13/021—Gearings
    • 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
    • G01M13/02—Gearings; Transmission mechanisms
    • G01M13/028—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
    • G01M13/04—Bearings
    • 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
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • 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

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Acoustics & Sound (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Signal Processing (AREA)
  • General Engineering & Computer Science (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

The invention relates to the technical field of signal processing of mechanical equipment, in particular to a method for diagnosing gray faults of key rotating parts of a signal frequency domain characteristic driving vehicle. According to the characteristic that the frequency domain characteristics of vibration signals caused by different faults are different, analyzing the vibration signals of the vehicle rotating part or envelope signals of the vibration signals by using the frequency response of Q-switched wavelet transform, and constructing a characteristic vector taking the energy ratio of each sub-band of the signals in different frequency bands as an element; aiming at the problem that when two groups of intersected sequences are processed by gray approximate association degrees, the two sequences have different change trends but the association degree is too large due to small original accumulation difference, a gray absolute approximate association degree model is provided; under the drive of the constructed frequency domain feature vector, the gray absolute proximity correlation degree of the frequency domain feature vector and the standard pattern is calculated to identify the fault state of the key rotating part of the vehicle. Compared with other vehicle key rotating part fault diagnosis methods, the method has the advantages of theoretical support and high algorithm efficiency.

Description

Method for diagnosing gray fault of key rotating part of signal frequency domain characteristic driven vehicle
Technical Field
The invention relates to a fault diagnosis method, in particular to a gray fault diagnosis method for a key rotating part of a signal frequency domain characteristic driven vehicle.
Background
The fault diagnosis technology is a key technology for improving the effectiveness of mechanical equipment, and determines the health state of the equipment through detection, extraction and identification of the operation state information of the mechanical equipment. As vehicles play an increasingly important role in transportation and industrial production, failure and damage to vehicles, particularly critical rotating components thereof at high speeds and high loads, can cause serious disasters and losses. Therefore, fault diagnosis of critical rotating components of a vehicle plays a crucial role in the reliability and safety of vehicle operation. However, most of the existing fault diagnosis methods need massive experimental data, lack of mechanism analysis of vibration response, have large influence on classification results and precision due to adjustment of parameters, and have high calculation force requirements, complex model design and the like. Therefore, a fault diagnosis method which has low requirement on the number of samples, does not require data to have a typical distribution rule, has a simple algorithm and a high calculation speed and is suitable for small samples is urgently needed at present.
Disclosure of Invention
The invention aims to solve the defects, and according to the characteristic that the frequency domain characteristics of vibration signals caused by different faults are different, the vibration signals of a vehicle rotating part or envelope signals of the vibration signals are analyzed by using the frequency response of Q-switched wavelet transform, and characteristic vectors taking the energy ratio of the signals in sub-bands of different frequency bands as elements are constructed; aiming at the problem that when two groups of intersected sequences are processed by gray approximate association degrees, the two sequences have different change trends but the association degree is too large due to small original accumulation difference, a gray absolute approximate association degree model is provided; under the drive of the constructed frequency domain feature vector, the gray absolute proximity correlation degree of the frequency domain feature vector and the standard pattern is calculated to identify the fault state of the key rotating part of the vehicle. A grey fault diagnosis method for key rotating parts of a signal frequency domain characteristic driving vehicle is provided.
In order to overcome the defects in the background art, the technical scheme adopted by the invention for solving the technical problems is as follows: a gray fault diagnosis method for a key rotating component of a vehicle driven by signal frequency domain characteristics is characterized by comprising the following steps: the method comprises the following steps:
step 1: and collecting vibration signals of key rotating parts of the vehicle in different running states. And selecting proper Q-switched wavelet parameters, and performing frequency domain response analysis on the acquired vibration signals or the envelopes thereof to obtain sub-band information under each frequency band. The frequency response of the Q-switched wavelet transform has a division effect on the frequency band of the signal from coarse to fine, and the adjustment of the frequency band division resolution can be realized by adjusting related parameters. Therefore, the step adopts Q-switched wavelet transform to decompose the vibration signal into sub-band signals, and realizes the division of signal frequency bands.
Step 2: and calculating the energy of each sub-band, and constructing a feature vector capable of mapping the health state of the part by taking the energy ratio of each sub-band as an element. Then, the mean value of the energy ratio distribution of any 5 samples in each state is taken as a standard mode, so as to represent the frequency domain characteristics in each healthy state.
And step 3: and (3) inputting the feature vector constructed by the sample to be recognized into the gray absolute proximity correlation degree model, and calculating the gray absolute proximity correlation degree between the sample to be recognized and each standard mode obtained in the step (2). And the standard mode with the maximum degree of correlation with the sample to be identified is the running state of the sample to be identified relative to the rotating component, so that the fault diagnosis of the key rotating component of the vehicle is realized.
The step 1 comprises the following steps:
step 1: and performing frequency domain response analysis on the acquired vibration signals or the envelopes thereof by adopting Q-switched wavelet transform to obtain sub-band information under each frequency band. The Q-switched wavelet transform is a wavelet transform method designed from the aspect of frequency domain filtering, and relies on a dual-channel filter bank comprising a decomposition filter and a reconstruction filter to realize the decomposition and reconstruction of signals through iterative operation. Wherein, the low-pass and high-pass scale factors alpha and beta are related to the quality factor Q and the redundancy factor r of the important parameters of Q-switched wavelet transform, and can be expressed as
Figure BDA0003301284100000021
And the parameters α and β also need to satisfy the following conditions:
0＜α＜1,0＜β≤1,α+β＞1 (2)
frequency response function H of low-pass and high-pass filters0(omega) and H1(ω) is:
Figure BDA0003301284100000022
Figure BDA0003301284100000023
in the formula, omega is angular frequency; theta (omega) is a frequency response function with a second order moment of disappearance,
Figure BDA0003301284100000024
|ω|≤π。
thus, given a quality factor Q and a redundancy factor r, a unique frequency response is obtained, with the frequency response range for the jth sub-band being [ (1-. beta.) αj-1π,αj-1π]The bandwidth and the center frequency of the jth sub-band are respectively
Figure BDA0003301284100000031
Figure BDA0003301284100000032
In the formula (f)sIs the sampling frequency. I.e. the order of the sub-bandsThe larger the number j, the smaller the bandwidth and center frequency of the frequency response. The number of decomposition layers J cannot be infinite, and given a quality factor Q and a redundancy factor r, the maximum number of decomposition layers JmaxIs composed of
Figure BDA0003301284100000033
Wherein, N is the length of the signal,
Figure BDA0003301284100000034
to round the symbol down.
The method is characterized in that the method comprises the steps of analyzing vibration response frequency domain characteristics of different fault types by adopting Q-switched wavelet transformation, and obtaining sub-band signals of information related to the health state of key rotating parts of the vehicle in different frequency bands.
The step 2 comprises the following steps:
step 2.1: after the frequency domain characteristics are obtained by adopting Q-switched wavelet transform to the signals, the energy E of each sub-band is calculatedj：
Figure BDA0003301284100000035
Where J ═ 1,2, …, J denotes subband ordinal; w is ajRepresenting the wavelet coefficients of the jth sub-band.
Step 2.2: calculating the ratio Tj of each sub-band energy to the total signal energy:
Figure BDA0003301284100000036
in the formula, EjRepresents the energy of the jth sub-band; t isjRepresenting the fraction of energy occupied by the jth subband. Signal frequency domain characteristic vector T ═ T for reflecting health state of parts is constructed by taking energy distribution of sub-bands as elements1,T2,…,TJ]。
Step 2.3: and constructing a signal frequency domain feature vector for each working condition sample according to the steps. Then, the mean value of the energy ratio distribution of any 5 samples from each state is taken as a standard mode, so as to represent the frequency domain characteristics in each healthy state.
In general, in steps 1 and 2, feature vectors with different frequency domain features and corresponding to the health states of the key rotating parts one by one are obtained by adjusting Q wavelet energy distribution, and standard modes in all the health states are constructed. Based on the method, the state monitoring and fault diagnosis of the key rotating parts of the vehicle can be realized through the pattern recognition method.
The step 3 comprises the following steps:
step 3.1: in order to make up the deficiency that the traditional gray proximity correlation degree is identified in an energy distribution angle, the extracted gray absolute proximity correlation degree model is adopted for pattern identification, and the definition is as follows:
i. j is a sequence label, and a system behavior sequence is set
Xi＝(xi(1),xi(2),…,xi(n)) (10)
Xj＝(xj(1),xj(2),…,xj(n)) (11)
Order to
Figure BDA0003301284100000041
Wherein k is 1,2, …, n. Then call
Figure BDA0003301284100000042
Grey absolute proximity correlation. The gray absolute proximity correlation model satisfies the gray correlation axiom and has the following properties:
a、0<ρijless than or equal to 1. According to the formula (12), Δ XijGreater than or equal to 0, and is proved by substituting formula (13).
b、ρ ii1. According to the formula (12), Δ XiiWhen the ratio is 0, the syndrome can be obtained.
c、ρij＝ρji. According to the formula (12), Δ Xij＝△XjiObtain the syndrome.
d、ρijNot only with XiAnd XjIs also related to its spatial position, i.e. translation will cause the correlation value to change;
e、Xiand XjThe closer together, ρijThe larger and vice versa.
Therefore, the grey absolute correlation degree can be used for calculating the correlation degree of the sample to be recognized and various state standard patterns, and thus the classification recognition is realized.
Step 3.2: and inputting the feature vector constructed by the sample to be recognized into the gray absolute proximity correlation degree model, and calculating the gray absolute proximity correlation degree between the sample to be recognized and each standard mode. And the standard mode with the maximum degree of correlation with the sample to be identified is the running state of the sample to be identified relative to the rotating component, so that the fault diagnosis of the key rotating component of the vehicle is realized.
The relevance model solves the problem that when the traditional grey proximity relevance model intersects with the fold lines formed by two groups of sequences, although the variation trends of the two fold lines are possibly completely different, the original accumulative difference is possibly very small, so that the relevance is too large, and the proximity of the two groups of sequences cannot be truly reflected. Meanwhile, the algorithm is simple, the calculation efficiency is high, the requirement on the sample is low compared with other intelligent classification methods, and the method is suitable for small sample pattern recognition. Through the steps, the method can accurately identify the operation state and the fault type of the key rotating part of the vehicle.
The invention has the beneficial effects that:
1. the invention firstly adopts Q-switched wavelet energy distribution to carry out feature extraction. When the mechanical equipment fails, the energy and the energy distribution of each frequency band of the vibration signal change, and the energy of each frequency band signal contains abundant failure information, so that the characteristic information can be extracted based on the energy distribution. In order to extract the energy distribution and characteristic information of the signal in the frequency domain, the signal needs to be decomposed. The Q-switched wavelet transform is a signal processing method capable of dividing high-frequency and low-frequency parts from coarse to fine, can finely depict the energy distribution of the low-frequency part of a signal, and can realize the adjustment of frequency band division resolution by adjusting related parameters, and key rotating parts of a vehicle, such as a bearing or a gear, are mainly reflected in the low-frequency part by fault characteristic frequency and fault-related energy distribution on a frequency domain. Therefore, the Q-switched wavelet energy distribution strategy adopted by the invention accurately extracts the state information contained in the vibration signal.
2. The grey absolute proximity relevance model effectively solves the problems that when two series of broken lines of the traditional grey proximity relevance are intersected, although the change trends of the two broken lines are possibly completely different, the original accumulation difference is possibly very small, the relevance is too large, and the proximity degree of the two series cannot be truly reflected, and can be more accurately evaluated. Based on the method, the gray fault diagnosis method driven by the signal frequency domain characteristics can successfully identify the working state and the fault type of the key rotating parts of the vehicle, has the advantages of low sample requirement, high calculation efficiency and the like compared with other intelligent classification methods, and provides a new method for the fault diagnosis of the key rotating parts of the vehicle.
Drawings
The invention is further illustrated with reference to the following figures and examples.
FIG. 1 is a schematic diagram of the construction of the test apparatus of the present invention;
FIG. 2 is a schematic diagram of a frequency domain feature-driven gray fault diagnosis algorithm;
FIG. 3 is a schematic diagram of the structure of a Q-switched wavelet transform frequency response with varying Q-factors;
FIG. 4 is a schematic diagram of the structure of a Q-switched wavelet transform frequency response with varying Q-factors;
FIG. 5 is a schematic diagram of the structure of a Q-switched wavelet transform frequency response with varying Q-factors;
FIG. 6 is a schematic diagram of the structure of the Q-switched wavelet transform frequency response with varying r-factor;
FIG. 7 is a schematic diagram of the structure of the Q-switched wavelet transform frequency response with varying r-factor;
FIG. 8 is a schematic diagram of the structure of the Q-switched wavelet transform frequency response with varying r-factor;
FIG. 9 is a schematic diagram of a standard mode energy distribution map under normal condition of a bearing;
FIG. 10 is a schematic diagram of a standard mode energy distribution map under a bearing inner race fault condition;
FIG. 11 is a schematic structural diagram of a standard pattern energy distribution diagram under a bearing outer ring fault condition;
FIG. 12 is a schematic structural diagram of a standard mode energy distribution diagram under a fault condition of a bearing rolling element;
FIG. 13 is a schematic structural diagram of a normal state sample identification result to be measured of a bearing under the condition of gray absolute proximity correlation;
FIG. 14 is a schematic structural diagram of a fault sample identification result of an inner ring to be detected of a bearing under the condition of gray absolute proximity correlation;
FIG. 15 is a schematic structural diagram of a fault sample identification result of an outer ring to be tested of a bearing under the condition of gray absolute proximity correlation;
FIG. 16 is a schematic structural diagram of a fault sample identification result of a rolling element to be tested of a bearing under the condition of gray absolute proximity correlation;
FIG. 17 is a schematic structural diagram of a normal state sample identification result to be measured of a bearing under the condition of adopting a gray approach correlation degree;
FIG. 18 is a schematic structural diagram of a fault sample identification result of an inner ring to be detected of a bearing under the condition of adopting gray approximate correlation;
FIG. 19 is a schematic structural diagram of a fault sample identification result of an outer ring to be tested of a bearing under the condition of adopting gray approximate correlation;
FIG. 20 is a structural diagram of a fault sample identification result of a rolling element to be tested of a bearing under the condition of adopting a gray approximate correlation degree;
FIG. 21 is a structural diagram of recognition results of the present invention method and other conventional methods in different training set numbers.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the following describes the present invention in detail with reference to actual experimental data:
the experimental data come from the train wheel set bearing test bench, and the test device is shown in figure 1. The inner ring of the bearing is connected with the main shaft and is driven by a driving motor, and the outer ring is fixed and connected with a radial load loading mechanism. Grooves with the width of about 0.18mm and the depth of about 1mm are respectively machined on the inner ring, the outer ring and the rolling body of the bearing by a wire cutting method, and the bearing is NJ (P)3226X 1. In the experiment, the rotating frequency is 10Hz, the characteristic frequency of the fault of the inner ring is 81.8Hz, the characteristic frequency of the fault of the outer ring is 58.2Hz and the characteristic frequency of the fault of the rolling body is 57.7 Hz. The sampling frequency is 5000Hz, the signal length is N which is 5000 points, and the duration t is 1 s. Four working conditions of normal bearing, inner ring fault, outer ring fault and rolling body fault of the train wheel set are respectively sampled, and 30 groups of data are obtained. Randomly selecting 5 groups of data, enabling the average value of the constructed feature vectors to be used as a standard mode, and using the remaining 25 groups of data as samples to be detected.
As shown in fig. 2, the present invention comprises the steps of:
step 1: and collecting vibration signals of key rotating parts of the vehicle in different running states. And selecting proper Q-switched wavelet parameters, and performing frequency domain response analysis on the acquired vibration signals or the envelopes thereof to obtain sub-band information under each frequency band. The frequency response of the Q-switched wavelet transform has a division effect on the frequency band of the signal from coarse to fine, and the adjustment of the frequency band division resolution can be realized by adjusting related parameters. Therefore, the step adopts Q-switched wavelet transform to decompose the vibration signal into sub-band signals, and realizes the division of signal frequency bands.
Step 2: and calculating the energy of each sub-band, and constructing a feature vector capable of mapping the health state of the part by taking the energy ratio of each sub-band as an element. Then, the mean value of the energy ratio distribution of any 5 samples in each state is taken as a standard mode, so as to represent the frequency domain characteristics in each healthy state.
And step 3: and (3) inputting the feature vector constructed by the sample to be recognized into the gray absolute proximity correlation degree model, and calculating the gray absolute proximity correlation degree between the sample to be recognized and each standard mode obtained in the step (2). And the standard mode with the maximum degree of correlation with the sample to be identified is the running state of the sample to be identified relative to the rotating component, so that the fault diagnosis of the key rotating component of the vehicle is realized.
Further, the step 1 specifically includes the following steps:
in step 1, the frequency domain response analysis is carried out on the acquired vibration signals or the envelopes thereof by adopting Q-switched wavelet transform, and sub-band information under each frequency band is obtained. The Q-switched wavelet transform is a wavelet transform method designed from the aspect of frequency domain filtering, and relies on a dual-channel filter bank comprising a decomposition filter and a reconstruction filter to realize the decomposition and reconstruction of signals through iterative operation. Wherein, the low-pass and high-pass scale factors alpha and beta are related to the quality factor Q and the redundancy factor r of the important parameters of Q-switched wavelet transform, and can be expressed as
Figure BDA0003301284100000071
And the parameters α and β also need to satisfy the following conditions:
0＜α＜1,0＜β≤1,α+β＞1 (2)
frequency response function H of low-pass and high-pass filters0(omega) and H1(ω) is:
Figure BDA0003301284100000072
Figure BDA0003301284100000073
in the formula, omega is angular frequency; theta (omega) is a frequency response function with a second order moment of disappearance,
Figure BDA0003301284100000074
|ω|≤π。
thus, given a quality factor Q and a redundancy factor r, a unique frequency response is obtained, with the frequency response range for the jth sub-band being [ (1-. beta.) αj-1π,αj-1π]The bandwidth and the center frequency of the jth sub-band are respectively
Figure BDA0003301284100000081
Figure BDA0003301284100000082
In the formula (f)sIs the sampling frequency. I.e., the larger the subband number j, the smaller the bandwidth and center frequency of the frequency response. The number of decomposition layers J cannot be infinite, and given a quality factor Q and a redundancy factor r, the maximum number of decomposition layers JmaxIs composed of
Figure BDA0003301284100000083
Wherein, N is the length of the signal,
Figure BDA0003301284100000084
to round the symbol down.
The quality factor Q is used to describe the degree of oscillation of the wavelet function and the redundancy factor r is used to represent the degree of overlap of the frequency responses. From equations (1) and (5), it can be deduced that when the redundancy factor r is fixed, an increase in the quality factor Q will narrow the bandwidth of the frequency response of each subband, which is advantageous for improving the frequency resolution. However, an increase in the quality factor Q leads to a decrease in the coverage frequency range, and therefore, a Q value matching the objective function should be selected by comprehensively considering a plurality of factors. Since the fault signature frequency of bearings and gears is often located in the low frequency band, it is often required to consider the value of Q in the case where both the useful frequency band is covered and the frequency resolution is high. As shown in fig. 3-5, the frequency response is given with a fixed r factor and an increased Q factor. It can be seen that as Q increases, the frequency resolution of the low band is improved, but the low frequency range that can be covered is reduced.
Furthermore, when the quality factor Q is fixed, an increase in the redundancy factor r will narrow the bandwidth of the frequency response of each sub-band and increase the frequency response resolution, as shown in fig. 6-8. The coverage frequency range is affected negligibly by the increase in r, but excessive redundancy greatly increases the computational cost, so the trade-off between frequency resolution and computational cost is taken into account to determine the value of r.
The number J of subbands is calculated by equation (7) by taking the maximum value of the frequency coverage range to be maximally obtained. Considering that the influence of the variation of the parameter Q factor and the redundancy r on the frequency response is basically consistent, in order to reduce the influence of the variable parameter, r is generally 3, and the influence of the Q factor on the frequency response is mainly analyzed. Table 1 gives the effect of each parameter on the bandwidth, coverage frequency range and computational cost of each sub-band frequency response.
TABLE 1 influence of parameters related to Q-switched wavelet transform
Figure BDA0003301284100000085
The method is characterized in that the method comprises the steps of analyzing vibration response frequency domain characteristics of different fault types by adopting Q-switched wavelet transformation, and obtaining sub-band signals of information related to the health state of key rotating parts of the vehicle in different frequency bands. According to the characteristics of the experimental signals, Q-switched wavelet transformation with parameters of Q-5, r-3 and J-45 is respectively carried out on envelope signals of different health states.
The step 2 specifically comprises the following steps:
step 2.1: after the frequency domain characteristics are obtained by adopting Q-switched wavelet transform to the signals, the energy E of each sub-band is calculatedj：
Figure BDA0003301284100000091
Where J ═ 1,2, …, J denotes subband ordinal; w is ajRepresenting the wavelet coefficients of the jth sub-band.
Step 2.2: calculating the ratio Tj of each sub-band energy to the total signal energy:
Figure BDA0003301284100000092
in the formula, EjRepresents the energy of the jth sub-band; t isjRepresenting the fraction of energy occupied by the jth subband. In sub-band energySignal frequency domain characteristic vector T ═ T for element construction reflecting part health state1,T2,…,TJ]。
Step 2.3: and constructing a signal frequency domain feature vector for each working condition sample according to the steps. Then, the mean value of the energy ratio distributions of any 5 samples in each state is taken as a standard mode, which represents the frequency domain characteristics in each healthy state, and standard mode energy distribution maps in four working conditions are obtained, as shown in fig. 9-12.
It can be noticed from the standard mode energy distribution diagram that the energy distributions corresponding to different operation states of the bearing are obviously different, and the bearing fault characteristic frequency is just located on the frequency band corresponding to the sub-band with the highest energy distribution. For example, the 28 th sub-band in the inner ring fault energy distribution diagram has the highest energy, and the inner ring fault characteristic frequency of 81.8Hz is just between the frequency band of 81.67Hz to 97.88Hz corresponding to the 28 th sub-band, so that the characteristics of the fault vibration signal are met, which is very beneficial to the identification of the fault type.
The step 3 specifically comprises the following steps:
step 3.1: in order to make up the deficiency that the traditional gray proximity correlation degree is identified in an energy distribution angle, the extracted gray absolute proximity correlation degree model is adopted for pattern identification, and the definition is as follows:
i. j is a sequence label, and a system behavior sequence is set
Xi＝(xi(1),xi(2),…,xi(n)) (10)
Xj＝(xj(1),xj(2),…,xj(n)) (11)
Order to
Figure BDA0003301284100000093
Wherein k is 1,2, …, n. Then call
Figure BDA0003301284100000094
Grey absolute proximity correlation.
And 3.2, inputting the feature vector constructed by the sample to be recognized into the gray absolute proximity correlation degree model, and calculating the gray absolute proximity correlation degree between the sample to be recognized and each standard mode. And the standard mode with the maximum degree of correlation with the sample to be identified is the running state of the sample to be identified relative to the rotating component, so that the fault diagnosis of the key rotating component of the vehicle is realized. In the experiment, the gray absolute proximity correlation degrees are respectively adopted for the state recognition of all 100 samples to be recognized, and partial recognition results are shown in fig. 13-16. Wherein the label "1" indicates a normal state, "2" indicates an inner ring failure, "3" indicates an outer ring failure, and "4" indicates a rolling body failure.
The recognition results are shown in fig. 13, and for the normal state samples, the correlation degree with the normal state standard pattern is about 0.9, while the correlation degree with other states is significantly lower, about 0.5-0.7. As further shown in fig. 14, for the inner ring fault sample, the degree of association with the inner ring fault standard mode is significantly higher than that of the other state standard modes. The identification result shows that the grey absolute proximity correlation model has a relatively ideal effect on the identification of the bearing health state, and the method can accurately and effectively identify the health state of the bearing.
In addition, for ease of comparison, FIGS. 17-20 show partial recognition results using a gray proximity correlation model. As shown in fig. 17, for the normal state sample, the correlation degree with the normal state standard pattern is about 0.98-1, and the correlation degree with other states is about 0.92-0.96, which can be classified correctly but the correlation degree is less. As shown in fig. 18, the correlation between the inner ring fault sample and the outer ring fault standard pattern is very close, and erroneous judgment is likely to be caused. In addition, the method is compared with the method provided by the text through three methods of decomposing the absolute proximity correlation degree with the gray based on empirical mode, converting the wavelet packet to the absolute proximity correlation degree with the gray based on the wavelet packet, and converting the Q-switched wavelet transform to the support vector machine. FIG. 21 shows the recognition results of different methods on different standard pattern samples, i.e. training set numbers; the number of samples to be identified, i.e., test sets, is 100.
As can be seen from the recognition results of fig. 6-8:
1. compared with the traditional gray proximity correlation degree, the gray absolute proximity correlation degree is obviously improved, and a more ideal identification effect is obtained; 2. the classification precision of the support vector machine is greatly influenced by the number of samples, and compared with the support vector machine, the gray correlation analysis has lower requirements on the number of samples and is suitable for solving the pattern recognition problem under smaller samples;
3. under the condition that the dimensionality of the constructed characteristic vectors is the same, the Q-switched wavelet transform obtains higher identification precision compared with wavelet packet transform, because under the condition that the dimensionality of the constructed characteristic vectors is the same, the Q-switched wavelet transform delicately describes the low frequency band of the signal;
4. compared with empirical mode decomposition, the Q-switched wavelet transform has higher identification precision, on one hand, the Q-switched wavelet transform is more finely divided into frequency bands, and on the other hand, the characteristic vector constructed by the empirical mode decomposition is difficult to be matched with the real frequency domain energy distribution of the signal in theory.
Therefore, the method can accurately identify the working state and the fault type of the key rotating part of the vehicle through the steps, has the advantages of low sample requirement, high calculation efficiency and the like compared with other intelligent classification methods, and provides a new method for the fault diagnosis of the key rotating part of the vehicle.

Claims (4)

1. A grey fault diagnosis method for a key rotating component of a signal frequency domain characteristic driving vehicle is characterized by comprising the following steps:
step 1, collecting vibration signals of a key rotating part of a vehicle in different running states, specifically selecting proper Q-switched wavelet parameters, and performing frequency domain response analysis on the collected vibration signals or envelopes thereof to obtain sub-band information under each frequency band;
step 2, calculating the energy of each sub-band, taking the energy ratio of each sub-band as an element, and constructing a feature vector capable of mapping the health state of the part, wherein the method specifically comprises the step of taking the energy ratio distribution mean value of any 5 samples of each state as a standard mode to represent the frequency domain feature of each state;
and 3, inputting the feature vector constructed by the sample to be recognized into the gray absolute proximity correlation degree model, calculating the gray absolute proximity correlation degree between the sample to be recognized and each standard mode obtained in the step 2, wherein the standard mode with the maximum correlation degree with the sample to be recognized is the running state of the sample to be recognized on the rotating part, and thus, the fault diagnosis of the key rotating part of the vehicle is realized.
2. The method for diagnosing the gray fault of the key rotating part of the signal frequency domain characteristic driven vehicle according to claim 1, wherein the Q-switched wavelet transform in the step 1 is a wavelet transform method designed from the aspect of frequency domain filtering, and realizes the decomposition and the reconstruction of the signal by iterative operation by relying on a set of two-channel filter banks containing decomposition and reconstruction filters. Wherein, the low-pass and high-pass scale factors alpha and beta are related to the quality factor Q and the redundancy factor r of the important parameters of Q-switched wavelet transform, and can be expressed as
Figure FDA0003301284090000011
And the parameters α and β also need to satisfy the following conditions:
0＜α＜1,0＜β≤1,α+β＞1 (2)
frequency response function H of low-pass and high-pass filters0(omega) and H1(ω) is:
Figure FDA0003301284090000012
Figure FDA0003301284090000013
in the formula, omega is angular frequency; theta (omega) is a toolThere is a frequency response function of the second order vanishing moment,
Figure FDA0003301284090000014
less than or equal to pi; thus, given a quality factor Q and a redundancy factor r, a unique frequency response is obtained, with the frequency response range for the jth sub-band being [ (1-. beta.) αj-1π,αj-1π]The bandwidth and the center frequency of the jth sub-band are respectively
Figure FDA0003301284090000021
Figure FDA0003301284090000022
In the formula (f)sIs the sampling frequency. I.e., the larger the subband number j, the smaller the bandwidth and center frequency of the frequency response. The number of decomposition layers J cannot be infinite, and given a quality factor Q and a redundancy factor r, the maximum number of decomposition layers JmaxIs composed of
Figure FDA0003301284090000023
Wherein, N is the length of the signal,
Figure FDA0003301284090000024
is a rounded-down symbol;
the method is characterized in that the method comprises the steps of analyzing vibration response frequency domain characteristics of different fault types by adopting Q-switched wavelet transformation, and obtaining sub-band signals of information related to the health state of key rotating parts of the vehicle in different frequency bands.
3. The method for diagnosing gray faults of key rotating parts of a signal frequency domain characteristic driven vehicle according to claim 1, wherein the step 2 comprises the following steps:
step 2.1: for signal acquisitionAfter the frequency domain characteristics are obtained by Q-switched wavelet transform, the energy E of each sub-band is calculatedj：
Figure FDA0003301284090000025
Where J ═ 1,2, …, J denotes subband ordinal; w is ajWavelet coefficients representing a jth sub-band;
step 2.2: calculating the ratio Tj of each sub-band energy to the total signal energy:
Figure FDA0003301284090000026
in the formula, EjRepresents the energy of the jth sub-band; t isjRepresenting the energy proportion occupied by the jth sub-band; signal frequency domain characteristic vector T ═ T for reflecting health state of parts is constructed by taking energy distribution of sub-bands as elements1,T2,…,TJ]；
Step 2.3: and (3) constructing a signal frequency domain feature vector by each working condition sample according to the steps, and then taking the energy ratio distribution average value of any 5 samples from each state as a standard mode to represent the frequency domain feature in each healthy state.
4. The method for diagnosing gray faults of key rotating parts of a signal frequency domain characteristic driven vehicle according to claim 1, wherein the step 3 comprises the following steps:
step 3.1, in order to make up the defect that the traditional gray proximity correlation degree is identified in an energy distribution angle, the extracted gray absolute proximity correlation degree model is adopted for pattern identification, and the definition is as follows:
i. j is a sequence label, and a system behavior sequence is set
Xi＝(xi(1),xi(2),…,xi(n)) (10)
Xj＝(xj(1),xj(2),…,xj(n)) (11)
Order to
Figure FDA0003301284090000031
Wherein, k is 1,2, …, n, then called
Figure FDA0003301284090000032
The gray absolute proximity correlation model satisfies the gray correlation axiom and has the following properties:
a、0<ρij1 or less, as shown by the formula (12), Δ XijNot less than 0, is proved by substituting formula (13),
b、ρiiwhen the formula (12) indicates that Δ X is 1iiObtaining the syndrome when the ratio is 0;
c、ρij＝ρjifrom the formula (12), Δ Xij＝△XjiObtaining the certificate;
d、ρijnot only with XiAnd XjIs also related to its spatial position, i.e. translation will cause the correlation value to change;
e、Xiand XjThe closer together, ρijThe larger the size, the smaller the size;
calculating the association degree of the sample to be identified and the standard modes in various states by utilizing the gray absolute association degree, thereby realizing classification identification; and 3.2, inputting the feature vector constructed by the sample to be recognized into the gray absolute proximity correlation degree model, calculating the gray absolute proximity correlation degree between the sample to be recognized and each standard mode, wherein the standard mode with the maximum correlation degree with the sample to be recognized is the running state of the sample to be recognized on the rotating part, and thus, the fault diagnosis of the key rotating part of the vehicle is realized.
CN202111191225.6A 2021-10-13 2021-10-13 Method for diagnosing gray fault of key rotating part of signal frequency domain characteristic driven vehicle Pending CN114091523A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111191225.6A CN114091523A (en) 2021-10-13 2021-10-13 Method for diagnosing gray fault of key rotating part of signal frequency domain characteristic driven vehicle

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111191225.6A CN114091523A (en) 2021-10-13 2021-10-13 Method for diagnosing gray fault of key rotating part of signal frequency domain characteristic driven vehicle

Publications (1)

Publication Number Publication Date
CN114091523A true CN114091523A (en) 2022-02-25

Family

ID=80296776

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111191225.6A Pending CN114091523A (en) 2021-10-13 2021-10-13 Method for diagnosing gray fault of key rotating part of signal frequency domain characteristic driven vehicle

Country Status (1)

Country Link
CN (1) CN114091523A (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060066290A1 (en) * 2004-09-27 2006-03-30 Reiner Hausdorf Method and apparatus for zero-mixing spectrum analysis with Hilbert transform
US20100023307A1 (en) * 2008-07-24 2010-01-28 University Of Cincinnati Methods for prognosing mechanical systems
CN105373700A (en) * 2015-10-30 2016-03-02 哈尔滨工程大学 Method for mechanical fault diagnosis based on information entropies and evidence theory
CN105865784A (en) * 2016-03-23 2016-08-17 大连理工大学 Rolling bearing detection method based on LMD (Local Mean Decomposition) and gray correlation
CN108398265A (en) * 2018-01-15 2018-08-14 上海电力学院 A kind of online fault detection method of rolling bearing
CN113095170A (en) * 2021-03-29 2021-07-09 天地（常州）自动化股份有限公司 Motor fault diagnosis method based on adjustable Q wavelet

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060066290A1 (en) * 2004-09-27 2006-03-30 Reiner Hausdorf Method and apparatus for zero-mixing spectrum analysis with Hilbert transform
US20100023307A1 (en) * 2008-07-24 2010-01-28 University Of Cincinnati Methods for prognosing mechanical systems
CN105373700A (en) * 2015-10-30 2016-03-02 哈尔滨工程大学 Method for mechanical fault diagnosis based on information entropies and evidence theory
CN105865784A (en) * 2016-03-23 2016-08-17 大连理工大学 Rolling bearing detection method based on LMD (Local Mean Decomposition) and gray correlation
CN108398265A (en) * 2018-01-15 2018-08-14 上海电力学院 A kind of online fault detection method of rolling bearing
CN113095170A (en) * 2021-03-29 2021-07-09 天地（常州）自动化股份有限公司 Motor fault diagnosis method based on adjustable Q wavelet

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
张彩芬 等: "一类新型绝对接近关联度分析模型及应用", 《运筹与管理》, 25 August 2013 (2013-08-25), pages 1 *

Similar Documents

Publication Publication Date Title
Yan et al. Wavelets for fault diagnosis of rotary machines: A review with applications
He et al. Tunable Q-factor wavelet transform denoising with neighboring coefficients and its application to rotating machinery fault diagnosis
Yang et al. Vibration feature extraction techniques for fault diagnosis of rotating machinery: a literature survey
CN111397896B (en) Fault diagnosis method and system for rotary machine and storage medium
Guo et al. Rolling bearing fault classification based on envelope spectrum and support vector machine
Osman et al. An enhanced Hilbert–Huang transform technique for bearing condition monitoring
Zhang et al. A new feature extraction approach using improved symbolic aggregate approximation for machinery intelligent diagnosis
CN108388908B (en) Rolling bearing impact fault diagnosis method based on K-SVD and sliding window noise reduction
Li et al. K-SVD-based WVD enhancement algorithm for planetary gearbox fault diagnosis under a CNN framework
Loutas et al. Utilising the wavelet transform in condition-based maintenance: A review with applications
CN107506710A (en) A kind of rolling bearing combined failure extracting method
CN114659790B (en) Identification method for bearing faults of variable-rotation-speed wind power high-speed shaft
CN109883706A (en) A kind of rolling bearing local damage Weak fault feature extracting method
Shuuji et al. Low-speed bearing fault diagnosis based on improved statistical filtering and convolutional neural network
Sahu et al. Fault diagnosis of rolling bearing based on an improved denoising technique using complete ensemble empirical mode decomposition and adaptive thresholding method
Lei et al. Application of a novel hybrid intelligent method to compound fault diagnosis of locomotive roller bearings
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
Hizarci et al. Vibration region analysis for condition monitoring of gearboxes using image processing and neural networks
Ambika et al. Vibration signal based condition monitoring of mechanical equipment with scattering transform
CN112733612A (en) Cross-domain rotating machinery fault diagnosis model establishing method and application thereof
Wang et al. CVRgram for demodulation band determination in bearing fault diagnosis under strong gear interference
Lee et al. Deep convolutional neural network with new training method and transfer learning for structural fault classification of vehicle instrument panel structure
CN107340133A (en) A kind of bearing condition monitoring method based on fitting Lifting Wavelet and higher order cumulants analysis
CN114091523A (en) Method for diagnosing gray fault of key rotating part of signal frequency domain characteristic driven vehicle
Shi et al. Rolling bearing feature frequency extraction using extreme average envelope decomposition

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