CN111783559A - Time-frequency domain joint fault diagnosis method based on vibration signals - Google Patents
Time-frequency domain joint fault diagnosis method based on vibration signals Download PDFInfo
- Publication number
- CN111783559A CN111783559A CN202010536722.4A CN202010536722A CN111783559A CN 111783559 A CN111783559 A CN 111783559A CN 202010536722 A CN202010536722 A CN 202010536722A CN 111783559 A CN111783559 A CN 111783559A
- Authority
- CN
- China
- Prior art keywords
- vector
- matrix
- threshold
- data
- fault
- 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.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- 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
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- General Engineering & Computer Science (AREA)
- Acoustics & Sound (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention belongs to the field of fault diagnosis methods of rotating machinery, and particularly relates to a time-frequency domain joint fault diagnosis method based on vibration signals, which comprises the steps of firstly, based on the characteristic that sampling of rotating signals has periodicity, representing sampling signals in a period as a multi-dimensional vector, representing sampling signals in a plurality of periods as a data matrix, solving characteristic values of the sampling signals by using an established covariance matrix, and establishing a characteristic value threshold vector for detecting real-time faults; and secondly, transforming the time domain matrix into a frequency domain matrix by utilizing wavelet packet transformation, further rewriting each row vector of the matrix into a matrix form based on certain periodicity, and solving a characteristic value to establish a corresponding threshold vector so as to judge which frequency band the fault occurs in, so that the positioning precision of the fault detection is improved, and more favorable support is provided for maintenance, management and the like of a system.
Description
Technical Field
The invention relates to the field of fault diagnosis methods for rotary machines, in particular to a time-frequency domain joint fault diagnosis method based on vibration signals.
Background
In the existing rotary machine fault diagnosis method, the effect is often poor because only the correlation based on one-dimensional vibration signals is utilized.
In the prior art, the most common one-dimensional vibration signal is collected from a rotary mechanical system, and time domain and frequency domain analysis is usually performed by methods such as wavelet transformation, neural network and the like. Wavelet transformation processes only the low-frequency part of a signal, which is unbalanced for frequency domain analysis, and the BP neural network has the defects of slow learning speed, easy falling into local minimum values and the like.
However, the above diagnostic method is still based on the analysis of the one-dimensional signal in the time domain, and only can improve the accuracy of the fault in the time domain. However, after determining the failure, one would want to know which frequency band the failure is carried on. If the method can be realized, the positioning precision of the fault is certainly improved, and a foundation is laid for further system maintenance and health management.
Disclosure of Invention
The invention aims to provide a time-frequency domain joint fault diagnosis method based on vibration signals.
The technical scheme for solving the technical problems is as follows:
a time-frequency domain joint fault diagnosis method based on vibration signals comprises the following steps:
1) the multivariate fault diagnosis method based on the time domain comprises the following steps: measuring vibration signal of rotary mechanical system, amplifying, transmitting and displaying the acquired vibration signal to obtain original vibration signal sequence
X(i)=[x(i)(1),x(i)(2),L,x(i)(L)]
Set forth below: i is 1,2, L, N, N +1, L, wherein i represents the ith group of data, L represents the length of each group of data, the first N groups are normal data and are used for training a diagnosis threshold value, and the rest groups of data are test data;
2) the sampled signal of the rotating mechanical system in one period is expressed as a vector, as follows: y is(i)=[x(i)(1),x(i)(2),Lx(i)(r)]TWherein r is the number of sampling points in a period;
the sampled signals over a number of cycles can be written in the form of a matrix as follows:
wherein, b ═ L/r ], i ═ 1,2, L, N;
order:
3) Determination of the threshold vector:
and recording the eigenvalue matrix as D:
SelectingThe vector of the threshold eigenvalue is obtained from the first N normal data, wherein α is a threshold coefficient used for adjusting the size of the selected threshold;
4) establishing an online fault discriminator:
7) the method is based on a fault diagnosis method on a frequency domain:
the wavelet packet operator matrix is recorded as W ∈ Rr×rAnd (3) carrying out wavelet packet transformation on the matrix form in the step (2) to obtain the following formula:
8) determining wavelet threshold values on each frequency range:
the above equation is recorded as the eigenvalue matrix form in step 3, and the eigenvalues are calculated:
Then, there is a matrix of feature values as in step 3:
recording:
selecting a threshold vector:
9) and judging the frequency band fault:
when there is new data Y(i)When coming, the feature vector is obtained based on the processes of the steps 7 to 8,
note that its deviation vector from the threshold vector is:
preferably, the total number L of sample points in step 2 is an integer multiple of r, and if not, the sample points are truncated or subjected to bit-filling processing.
Preferably, the total number b of sample points in step 8 is an integer multiple of l, and if not, the sample points are truncated or subjected to bit-filling processing.
The invention has the beneficial effects that: firstly, based on the characteristic that the sampling of a rotation signal has periodicity, a sampling signal in one period is expressed as a multi-dimensional vector, sampling signals in a plurality of periods are expressed in the form of a data matrix, a characteristic value of the sampling signal is solved by utilizing an established covariance matrix, and a threshold vector of the characteristic value is established for detecting real-time faults, so that whether the faults occur or not is judged in a time domain, a one-dimensional vibration signal is rewritten into the multi-dimensional signal, and whether the faults occur or not is judged by solving the threshold vector and establishing an online fault discriminator;
then, the time domain matrix is converted into a frequency domain matrix by wavelet packet conversion, each row vector of the matrix is further rewritten into a matrix form based on certain periodicity, and characteristic values are obtained again to establish corresponding threshold vectors so as to judge which frequency the fault occurs at, thereby judging the frequency band of the fault occurrence in the frequency domain. The time domain signal is transformed to the frequency domain by wavelet packet transformation, and a fault discriminator is established to discriminate which frequency band the fault occurs in by solving the threshold vector of the frequency band, so that people can know which frequency the fault occurs at, and the problems that the detection and diagnosis method is not good and the maintenance, management and the like of the system are inconvenient because the diagnosis method in the prior art can only know whether the fault exists in the time domain and can not know which frequency band the fault is carried on.
Detailed Description
The following examples are only for illustrating the technical solutions of the present invention more clearly, and therefore are only examples, and the protection scope of the present invention is not limited thereby.
A time-frequency domain joint fault diagnosis method based on vibration signals comprises the following steps:
1) the multivariate fault diagnosis method based on the time domain comprises the following steps: measuring vibration signals of a rotary mechanical system, amplifying, transmitting and displaying the acquired vibration signals, wherein the obtained original vibration signal sequence is as follows:wherein i represents the ith group of data, L represents the length of each group of data, the first N groups of data are normal data for training a diagnostic threshold, and the rest data groups are test data;
2) the sampled signal of the rotating mechanical system in one period is expressed as a vector, as follows: y is(i)=[x(i)(1),x(i)(2),Lx(i)(r)]TWherein r is the number of sampling points in a period;
the sampled signals over a number of cycles can be written in the form of a matrix as follows:
wherein, b ═ L/r ], i ═ 1,2, L, N;
order:
3) Determination of the threshold vector:
and recording the eigenvalue matrix as D:
SelectingThe vector of the threshold eigenvalue is obtained from the first N normal data, wherein α is a threshold coefficient used for adjusting the size of the selected threshold;
4) establishing an online fault discriminator:
7) the method is based on a fault diagnosis method on a frequency domain:
the wavelet packet operator matrix is recorded as W ∈ Rr×rAnd (3) carrying out wavelet packet transformation on the matrix form in the step (2) to obtain the following formula:
8) determining wavelet threshold values on each frequency range:
the above equation is recorded as the eigenvalue matrix form in step 3, and the eigenvalues are calculated:
Then, there is a matrix of feature values as in step 3:
recording:
selecting a threshold vector:
9) and judging the frequency band fault:
when there is new data Y(i)When coming, the feature vector is obtained based on the processes of the steps 7 to 8,
note that its deviation vector from the threshold vector is:
in the time-frequency domain joint fault diagnosis method based on the vibration signal, when in use, the judgment on the time domain is to judge whether the group data has faults or not, the process is as in the step 2 to the step 6, the judgment on the frequency domain is to judge which frequency band the faults occur in the fault data group, the process is as in the step 7 to the step 9,
firstly, as the process steps 2 to 6, a sampling signal in one period is expressed into a multidimensional vector, sampling signals in a plurality of periods are expressed into a data matrix form, the characteristic value of the sampling signal is obtained by utilizing the established covariance matrix, and a threshold vector of the characteristic value is established for detecting real-time faults, so that whether the faults occur or not is judged in a time domain, a one-dimensional vibration signal is rewritten into the multidimensional signal, and whether the faults occur or not is judged by obtaining the threshold vector and establishing an online fault discriminator;
then, as in the procedure from step 7 to step 9, the time domain matrix is transformed into a frequency domain matrix by wavelet packet transformation, each row vector of the matrix is further rewritten into a matrix form based on a certain periodicity, and the eigenvalue is obtained again to establish a corresponding threshold vector, thereby discriminating the frequency band in which the fault occurs in the frequency domain. Transforming the time domain signal to a frequency domain by utilizing wavelet packet transformation, and establishing a fault discriminator by solving a threshold vector of a frequency band to discriminate which frequency band a fault occurs;
thereby judging the frequency of the fault, improving the positioning precision of the fault detection, providing more favorable support for the maintenance, management and the like of the system,
therefore, people can know which frequency the fault occurs at, and the problems that the diagnosis method in the prior art can only know whether the fault exists in the time domain and cannot know which frequency band the fault is carried on, so that the detection diagnosis method is poor, and the system is inconvenient to maintain, manage and the like are solved.
Further, the total number L of sample points in step 2 is an integer multiple of r, and if not, the sample points are truncated or bit-complemented.
Further, the total number b of sample points in the step 8 is an integral multiple of l, and if not, the sample points are truncated or subjected to bit-filling processing.
Furthermore, the invention can know which frequency the fault occurs in through the method, and can combine deep learning to solve the problem when a plurality of faults are in a plurality of frequency bands.
Further, in order to verify the aforementioned method, a failure needs to be added to the test data. Since the fault on the frequency band is needed, firstly, the fault data group to be added is rewritten into a matrix form, and a wavelet packet matrix is obtained through wavelet packet transformation, namely, different frequency bands are obtained. And adding the fault into different frequency bands, carrying out wavelet packet reconstruction on the fault, and processing the reconstructed data serving as an original signal. Here we choose to add the fault as 0.8 times the maximum deviation of the relevant frequency band in the normal data.
In this embodiment: the simulation adopts data of a bearing data center website of the university of western storage, 76 groups of data are extracted from data with the rotating speed of 1797rpm and the load of 0HP, the first 60 groups are used as training data, and the last 16 groups are used as test data;
the simulation results obtained by performing 18 sets of tests on different data sets with faults added on different frequency bands are shown in tables 1 and 2:
TABLE 1 time domain discrimination results
Fault data set | Time domain discrimination result |
67 | 67 |
61 | 61 |
75 | 75 |
67 | 67 |
61 | 61 |
75 | 75 |
70 | 70 |
72 | 72 |
69 | 69 |
68 | 68 |
67 | 67 |
61 | 61 |
75 | 75 |
70 | 70 |
72 | 72 |
69 | 69 |
68 | 68 |
64 | None |
TABLE 2 frequency domain discrimination results
Frequency band of faults | Frequency domain discrimination result |
5 | 5 |
2 | 2 |
5 | 4 5 |
2 5 | 2 5 |
5 7 | 5 7 |
5 7 | 4 5 7 |
3 8 | 3 8 |
1 8 | 1 8 |
2 4 | 2 4 8 |
4 7 | 4 7 |
1 4 5 | 1 4 5 |
1 5 7 | 1 5 7 |
2 5 7 | 2 4 5 7 |
3 6 8 | 3 6 8 |
1 4 8 | 1 4 8 |
2 4 8 | 2 4 8 |
2 4 7 | 2 4 7 |
None | None |
From the simulation results, the method can not only detect which group of data fails, but also detect the type of the failed frequency band. As in the 17 th test in the table, faults were added to the low, mid and high frequency bands in the 68 th set of data, and the test results were all good. However, the test results of groups 3, 6, 9 and 13 show that the 4 th frequency band has faults all the time, and the faults are false alarms through simulation verification, so the effectiveness of the method is verified through simulation.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; the modifications and the substitutions do not cause the essence of the corresponding technical solutions to depart from the scope of the technical solutions of the embodiments of the present invention, and the corresponding technical solutions are all covered in the claims and the specification of the present invention.
Claims (3)
1. A time-frequency domain joint fault diagnosis method based on vibration signals is characterized by comprising the following steps:
1) the multivariate fault diagnosis method based on the time domain comprises the following steps: measuring vibration signals of a rotary mechanical system, amplifying, transmitting and displaying the acquired vibration signals, wherein the obtained original vibration signal sequence is as follows:wherein i represents the ith group of data, L represents the length of each group of data, the first N groups of data are normal data for training a diagnostic threshold, and the rest data groups are test data;
2) the sampled signal of the rotating mechanical system in one period is expressed as a vector, as follows: y is(i)=[x(i)(1),x(i)(2),Lx(i)(r)]TWherein r is the number of sampling points in a period;
the sampled signals over a number of cycles can be written in the form of a matrix as follows:
wherein, b ═ L/r ], i ═ 1,2, L, N;
order:
3) Determination of the threshold vector:
and recording the eigenvalue matrix as D:
SelectingThe vector of the threshold eigenvalue is obtained from the first N normal data, wherein α is a threshold coefficient used for adjusting the size of the selected threshold;
4) establishing an online fault discriminator:
7) the method is based on a fault diagnosis method on a frequency domain:
the wavelet packet operator matrix is recorded as W ∈ Rr×rAnd C, performing wavelet packet transformation on the matrix form in the step 2 to obtain the following formula:
8) determining wavelet threshold values on each frequency range:
the above equation is recorded as the eigenvalue matrix form in step 3, and the eigenvalues are calculated:
Then, there is a matrix of feature values as in step 3:
recording:
selecting a threshold vector:
9) and judging the frequency band fault:
when there is new data Y(i)When coming, the feature vector is obtained based on the processes of the steps 7 to 8,
note that its deviation vector from the threshold vector is:
2. the method according to claim 1, wherein the total number L of sample points in step 2 is an integer multiple of r, and if not, the method performs truncation or bit-filling processing on the sample points.
3. The method according to claim 1, wherein the total number b of the sample points in the step 8 is an integer multiple of l, and if not, the total number b of the sample points is truncated or subjected to bit-filling processing.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010536722.4A CN111783559A (en) | 2020-06-12 | 2020-06-12 | Time-frequency domain joint fault diagnosis method based on vibration signals |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010536722.4A CN111783559A (en) | 2020-06-12 | 2020-06-12 | Time-frequency domain joint fault diagnosis method based on vibration signals |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111783559A true CN111783559A (en) | 2020-10-16 |
Family
ID=72756286
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010536722.4A Withdrawn CN111783559A (en) | 2020-06-12 | 2020-06-12 | Time-frequency domain joint fault diagnosis method based on vibration signals |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111783559A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113177537A (en) * | 2021-06-29 | 2021-07-27 | 湖北博华自动化系统工程有限公司 | Fault diagnosis method and system for rotary mechanical equipment |
-
2020
- 2020-06-12 CN CN202010536722.4A patent/CN111783559A/en not_active Withdrawn
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113177537A (en) * | 2021-06-29 | 2021-07-27 | 湖北博华自动化系统工程有限公司 | Fault diagnosis method and system for rotary mechanical equipment |
CN113177537B (en) * | 2021-06-29 | 2021-09-17 | 湖北博华自动化系统工程有限公司 | Fault diagnosis method and system for rotary mechanical equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109858104B (en) | Rolling bearing health assessment and fault diagnosis method and monitoring system | |
CN108510153B (en) | Multi-working-condition rotary machine fault diagnosis method | |
CN109916628B (en) | Rolling bearing fault diagnosis method based on improved multi-scale amplitude perception permutation entropy | |
Lei et al. | EEMD method and WNN for fault diagnosis of locomotive roller bearings | |
CN110398647B (en) | Transformer state monitoring method | |
WO2019061006A1 (en) | Bearing failure diagnosis method and device, readable storage medium, and electronic device | |
Chen et al. | Compound fault identification of rolling element bearing based on adaptive resonant frequency band extraction | |
CN104596766B (en) | Early fault determining method and device for bearing | |
CN112179691A (en) | Mechanical equipment running state abnormity detection system and method based on counterstudy strategy | |
CN112147444B (en) | Power transformer working state monitoring method and system | |
CN117272210A (en) | Method and system for detecting abnormal hidden danger data of building construction | |
CN110426191A (en) | A kind of method for diagnosing faults of anti-interference rotating machinery | |
CN113086799A (en) | Elevator fault detection and early warning method based on image recognition | |
CN109139390A (en) | A kind of fan blade fault recognition method based on acoustical signal feature database | |
CN111783559A (en) | Time-frequency domain joint fault diagnosis method based on vibration signals | |
CN114323642A (en) | Wind turbine generator vibration data processing system and data dilution method | |
CN114486263B (en) | Noise reduction demodulation method for vibration signal of rolling bearing of rotary machine | |
Wang et al. | Multi-domain extreme learning machine for bearing failure detection based on variational modal decomposition and approximate cyclic correntropy | |
CN113139430A (en) | Acoustic signal analysis method for fault detection, and method and device for fault detection | |
CN113466616A (en) | Method and device for quickly positioning cable fault point | |
CN114781466B (en) | Fault diagnosis method and system based on harmonic fundamental frequency of rotary mechanical vibration signal | |
Madhukar et al. | A Machine Learning Based Methodology for Fault Diagnosis in Rotating Machine | |
CN112049787A (en) | Non-contact water pump fault detection method and system | |
CN115048960A (en) | Equipment state detection method | |
CN113487141B (en) | Bearing state evaluation method based on multisource information clustering fusion |
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 | ||
WW01 | Invention patent application withdrawn after publication | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20201016 |