CN111076934A - Method for diagnosing potential fault of bearing based on S transformation - Google Patents

Method for diagnosing potential fault of bearing based on S transformation Download PDF

Info

Publication number
CN111076934A
CN111076934A CN201911347912.5A CN201911347912A CN111076934A CN 111076934 A CN111076934 A CN 111076934A CN 201911347912 A CN201911347912 A CN 201911347912A CN 111076934 A CN111076934 A CN 111076934A
Authority
CN
China
Prior art keywords
fault
bearing
data set
transformation
time domain
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
CN201911347912.5A
Other languages
Chinese (zh)
Inventor
王秀礼
李志国
朱荣生
付强
赵媛媛
徐伟
林彬
蒋夏飞
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 YUANQUAN PUMP INDUSTRY Co.,Ltd.
Original Assignee
Jiangsu 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 Jiangsu University filed Critical Jiangsu University
Priority to CN201911347912.5A priority Critical patent/CN111076934A/en
Publication of CN111076934A publication Critical patent/CN111076934A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00—Testing of machine parts
    • G01M13/04—Bearings
    • G01M13/045—Acoustic or vibration analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Acoustics & Sound (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
  • Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)

Abstract

The invention provides a method for diagnosing potential faults of a bearing based on S transformation, which comprises the following steps: collecting vibration signals of bearings with different fault types, and filtering the vibration signals by utilizing wavelet denoising; s transformation is carried out on the filtered signals to obtain a feature vector set of a time domain and a frequency domain of the jth fault type; using the obtained time domain and frequency domain feature vectors of the jth fault type as the input of an SAE deep learning model to perform data dimension reduction processing to obtain a data set Xj(ii) a Data set X of different fault typesjForming a sample data set, using the sample data set and the fault type as the input of a support vector machine, and training to obtain a bearing fault identification model; and identifying the bearing vibration signal with unknown fault by using a bearing fault identification model. The invention decomposes the fault information into a two-dimensional matrix containing time domain and frequency domain information through S transformation time-frequency analysis, which is beneficial to observing the change of signals when the fault is nascent.

Description

Method for diagnosing potential fault of bearing based on S transformation
Technical Field
The invention relates to the field of bearing fault analysis, in particular to a method for diagnosing potential faults of a bearing based on S transformation.
Background
Machines with rotating parts wear out over time due to prolonged use, even if the machine is maintained and/or repaired regularly. Some machines are not monitored for wear during use, but are only inspected between uses. For example, turbochargers in certain vehicles (e.g., locomotives) may have no instrumentation to monitor vibration, bearing temperature, or rotor response during turbocharger operation. Excessive wear of the bearings can lead to rotor instability, high displacement and eventual friction, which can lead to turbocharger failure. High imbalance caused by Foreign Object Damage (FOD) or excessive deposits on turbine blades can also lead to high shaft motion, bearing wear, and eventual failure due to rotational contact with stationary parts.
The prior art discloses a plunger pump fault diagnosis system based on dual-class feature fusion diagnosis, which converts a vibration signal of a pump into an electric signal through an acceleration sensor, and extracts two types of features of a wavelet packet relative energy spectrum and a wavelet packet relative feature entropy of the signal in a dual-class feature extraction mode to perform fault diagnosis. The actual operation of the system needs to connect the acceleration sensor with the pump body, and the characteristics of the vibration signals have great relation with the connection position, so that the operation at a plurality of positions of the pump body is needed to accurately monitor a certain fault, the process is complicated and time-consuming, and professional operators are needed to monitor under most conditions.
The patent in the prior art discloses a bearing early fault identification method based on a long-time and short-time memory cyclic neural network. And extracting common time domain characteristics after the bearing vibration signals are collected, constructing a characteristic data set by using the time domain characteristics and the entropy characteristics, and training the LSTM recurrent neural network by using the characteristic data set as a training sample. And identifying the fault occurrence time through the trained LSTM recurrent neural network. The method combines the traditional characteristics and entropy characteristics of the vibration signal, and accurately reflects the current state of the bearing under the condition of ensuring the physical significance of the vibration characteristic quantity. But the misjudgment caused by the interference signal can not be effectively distinguished.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a method for diagnosing the potential fault of the bearing based on S transformation, which decomposes fault information into a two-dimensional matrix containing time domain and frequency domain information through S transformation time-frequency analysis and is beneficial to observing the change of signals when the fault is initially generated.
The present invention achieves the above-described object by the following technical means.
A method for diagnosing latent faults of a bearing based on S transformation comprises the following steps:
acquiring vibration signals x of bearings with different fault typesj(t) denoising the vibration signal x using a waveletj(t) filtering to obtain xj(m), wherein j is a fault category;
for filtered xj(m) carrying out S transformation on the signal to obtain a feature vector set of a time domain and a frequency domain of the jth fault type;
using the obtained time domain and frequency domain feature vectors of the jth fault type as the input of an SAE deep learning model to perform data dimension reduction processing to obtain a data set Xj；
Data set X of different fault typesjForming a sample data set, using the sample data set and the fault type as the input of a support vector machine, and training to obtain a bearing fault identification model;
and identifying the bearing vibration signal with unknown fault by using a bearing fault identification model.
Further, the filtered x is filteredj(m) performing S-transform on the signal, the specific steps are as follows:
for signal xj(m) performing a discrete Fourier transform to Xj[k]：
Figure BDA0002333896520000021
In the formula, k is a frequency point of a frequency spectrum and takes the value of a natural number of 0,1, … and N-1; n is the number of sampling points;
x is to bej(m) and Xj[k]Performing S transformation, specifically as follows:
Figure BDA0002333896520000022
in the formula: t is a sampling period; n is the number of sampling points;
m is a time sampling point serial number and takes the value of a natural number of 0,1, … and N-1;
n is the serial number of the frequency sampling point, and the value is the natural number 0,1, …, N-1;
Figure BDA0002333896520000023
a set of matrix vectors in time domain and frequency domain for the jth fault type, where mT denotes a time domain sequence,
Figure BDA0002333896520000024
representing a sequence of frequencies.
Further, a set of matrix vectors of time domain and frequency domain of the jth fault type
Figure BDA0002333896520000025
As input to the SAE deep learning model, wherein: mT is the number of input layer elements of the SAE deep learning model,
Figure BDA0002333896520000026
the number of input samples as an SAE deep learning model.
Further, the support vector machine is LIBSVM; and obtaining a regularization parameter C and a Gaussian kernel parameter gamma of the constructed bearing fault identification model through cross validation grid optimization.
The invention has the beneficial effects that:
1. according to the method for diagnosing the potential fault of the bearing based on the S transformation, the fault information is decomposed into the two-dimensional matrix containing the time domain and the frequency domain information through the S transformation time-frequency analysis, and the method is beneficial to observing the change of signals when the fault is initiated.
2. According to the method for diagnosing the potential fault of the bearing based on the S transformation, disclosed by the invention, the data are subjected to dimensionality reduction through an SAE deep learning model, and a LIBSVM-based fault classification model is constructed, so that the method has a practical application value for identifying the fault of the bearing.
Drawings
FIG. 1 is a flow chart of a method for diagnosing a latent fault of a bearing based on an S-transform according to the present invention.
Fig. 2 is a comparison diagram before and after the wavelet denoising according to the present invention.
FIG. 3 is a SAE deep learning model according to the present invention.
FIG. 4 is a cross-validation grid optimization diagram according to the present invention.
Detailed Description
The invention will be further described with reference to the following figures and specific examples, but the scope of the invention is not limited thereto.
As shown in FIG. 1, the method for diagnosing latent faults of a bearing based on S transformation comprises the following steps:
acquiring vibration signals x of bearings with different fault typesj(t) denoising the vibration signal x using a waveletj(t) filtering to obtain xj(m), wherein j is a fault category;
for filtered xj(m) carrying out S transformation on the signal to obtain a feature vector set of a time domain and a frequency domain of the jth fault type;
using the obtained time domain and frequency domain feature vectors of the jth fault type as the input of an SAE deep learning model to perform data dimension reduction processing to obtain a data set Xj；
Data set X of different fault typesjForming a sample data set, using the sample data set and the fault type as the input of a support vector machine, and training to obtain a bearing fault identification model;
and identifying the bearing vibration signal with unknown fault by using a bearing fault identification model.
Example (b):
the bearing at the driving end of a certain motor is a deep grooveThe model of the ball bearing is 6324C3, the number of rollers N is 12, the inner diameter D is 120mm, the outer diameter D is 260mm, the contact angle β is 0 degree, the sampling frequency is 2000Hz, the motor speed 1450r/min is used for collecting vibration signals x for different faults of the bearing respectivelyj(t), j indicates the type of fault, where different faults include: normal state, cage failure, rolling element failure, outer ring failure, inner ring failure. x is the number of1(t) is a normal state, x2(t) cage failure, x3(t) failure of rolling elements, x4(t) outer ring failure and x5(t) inner ring failure.
Next, a detailed description will be made regarding a bearing inner race failure.
For the collected bearing inner ring fault signal x5(t) performing wavelet de-noising to obtain x5(m) as shown in FIG. 2.
For signal x5(m) performing a discrete Fourier transform to X5[k]：
Figure BDA0002333896520000041
In the formula, k is a frequency point of a frequency spectrum and takes the value of a natural number of 0,1, … and N-1; n is the number of sampling points;
x is to be5(m) and X5[k]Performing S transformation, specifically as follows:
Figure BDA0002333896520000042
in the formula: t is a sampling period; n is the number of sampling points;
m is a time sampling point serial number and takes the value of a natural number of 0,1, … and N-1;
n is the serial number of the frequency sampling point, and the value is the natural number 0,1, …, N-1;
Figure BDA0002333896520000043
a set of matrix vectors in time and frequency domains for the 5 th fault type, where mT denotes a time domain sequence,
Figure BDA0002333896520000044
representing a sequence of frequencies.
As shown in table 1, the bearing inner race fault data is subjected to the S transformation to obtain matrix partial data. In which Table 1 represents a time series mT in the horizontal direction and Table 1 represents a frequency series in the vertical direction
Figure BDA0002333896520000045
TABLE 1 partial data of inner ring fault after S transformation
Figure BDA0002333896520000046
Using the obtained feature vectors of time domain and frequency domain of the 5 th fault type as the input of an SAE deep learning model to perform data dimension reduction processing to obtain a data set X5(ii) a I.e., the time and frequency domain matrix vector sets for the 5 th fault type in Table 1
Figure BDA0002333896520000047
As input to the SAE deep learning model, wherein: mT is the number of input layer elements of the SAE deep learning model,
Figure BDA0002333896520000048
the number of input samples as an SAE deep learning model. FIG. 4 is the SAE deep learning model. Obtaining a data set X by carrying out dimensionality reduction processing on the self-contained software package of matlab5. Wherein the number of points of the input layer is
Figure BDA0002333896520000051
The number of hidden layers is 1000 and the number of output layers is 40.
And performing the above steps on the data of normal state, retainer fault, rolling element fault and outer ring fault, and forming data sets Xj with different fault types into a sample data set to obtain the sample data set shown in Table 2.
TABLE 2 partial data of data set
Figure BDA0002333896520000052
And training the sample data set and the corresponding fault category as input of the LIBSVM. The total number of samples used for each type of fault is 1000, the number of training samples is 800, and the number of testing samples is 200. The LIBSVM model is a matlab self-contained multi-classification support vector machine software package.
Building a bearing fault classifier through LIBSVM requires optimizing two parameters, a regularization parameter C gaussian kernel parameter γ. The invention utilizes the cross validation grid optimizing graph to carry out parameter optimizing. The principle is that C and gamma parameter combinations in a certain range are selected for model training, and finally the parameter with the highest resolution of the model can be used as the optimal parameter. FIG. 4 is a cross-validation grid optimization diagram. The horizontal axis represents different gamma parameters, the vertical axis represents different C parameters, and the middle square shows the model resolution under different C, gamma combinations. In this example C, γ is 0.001,0.01,0.1,1,10, 100. In this example, when C is 100, γ is 0.01, which is the highest resolution of the model, and is 97%. However, the values of C and γ are not limited to these 6 numbers, and may be adjusted according to the accuracy of the desired model.
And constructing bearing fault recognition models of different models, and forming a bearing fault database.
And identifying the bearing vibration signal of unknown fault by using the bearing fault identification model, and judging the fault type.
The present invention is not limited to the above-described embodiments, and any obvious improvements, substitutions or modifications can be made by those skilled in the art without departing from the spirit of the present invention.

Claims (4)

1. A method for diagnosing latent faults of a bearing based on S transformation is characterized by comprising the following steps:
acquiring vibration signals x of bearings with different fault typesj(t) denoising the vibration signal x using a waveletj(t) filtering to obtain xj(m) where j is a faultA category;
for filtered xj(m) carrying out S transformation on the signal to obtain a feature vector set of a time domain and a frequency domain of the jth fault type;
using the obtained time domain and frequency domain feature vectors of the jth fault type as the input of an SAE deep learning model to perform data dimension reduction processing to obtain a data set Xj；
Data set X of different fault typesjForming a sample data set, using the sample data set and the fault type as the input of a support vector machine, and training to obtain a bearing fault identification model;
and identifying the bearing vibration signal with unknown fault by using a bearing fault identification model.
2. The method for S-transform-based diagnosis of latent bearing failure according to claim 1, wherein the filtered x is filteredj(m) performing S-transform on the signal, the specific steps are as follows:
for signal xj(m) performing a discrete Fourier transform to Xj[k]：
Figure FDA0002333896510000011
In the formula, k is a frequency point of a frequency spectrum and takes the value of a natural number of 0,1, … and N-1; n is the number of sampling points;
x is to bej(m) and Xj[k]Performing S transformation, specifically as follows:
Figure FDA0002333896510000012
in the formula: t is a sampling period; n is the number of sampling points;
m is a time sampling point serial number and takes the value of a natural number of 0,1, … and N-1;
n is the serial number of the frequency sampling point, and the value is the natural number 0,1, …, N-1;
Figure FDA0002333896510000013
a set of matrix vectors in time domain and frequency domain for the jth fault type, where mT denotes a time domain sequence,
Figure FDA0002333896510000014
representing a sequence of frequencies.
3. The method for S-transform-based diagnosis of latent bearing failure according to claim 2, wherein the set of time domain and frequency domain matrix vectors for the jth failure type
Figure FDA0002333896510000015
As input to the SAE deep learning model, wherein: mT is the number of input layer elements of the SAE deep learning model,
Figure FDA0002333896510000016
the number of input samples as an SAE deep learning model.
4. The method for diagnosing latent bearing failure based on S-transform of claim 1, wherein the support vector machine is LIBSVM; and obtaining a regularization parameter C and a Gaussian kernel parameter gamma of the constructed bearing fault identification model through cross validation grid optimization.
CN201911347912.5A 2019-12-24 2019-12-24 Method for diagnosing potential fault of bearing based on S transformation Pending CN111076934A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911347912.5A CN111076934A (en) 2019-12-24 2019-12-24 Method for diagnosing potential fault of bearing based on S transformation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911347912.5A CN111076934A (en) 2019-12-24 2019-12-24 Method for diagnosing potential fault of bearing based on S transformation

Publications (1)

Publication Number Publication Date
CN111076934A true CN111076934A (en) 2020-04-28

Family

ID=70317324

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911347912.5A Pending CN111076934A (en) 2019-12-24 2019-12-24 Method for diagnosing potential fault of bearing based on S transformation

Country Status (1)

Country Link
CN (1) CN111076934A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111832428A (en) * 2020-06-23 2020-10-27 北京科技大学 Data enhancement method applied to strip breakage fault diagnosis of cold rolling mill
CN112232414A (en) * 2020-10-16 2021-01-15 广东石油化工学院 Triple concurrency fault analysis method based on X and Y dual-measurement-point spectrum data
CN112304611A (en) * 2020-09-14 2021-02-02 江苏大学 Deep learning-based bearing fault diagnosis method
CN115358277A (en) * 2022-10-09 2022-11-18 深圳市信润富联数字科技有限公司 Bearing fault diagnosis method, device, equipment and readable storage medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105447502A (en) * 2015-11-05 2016-03-30 国家电网公司 Transient power disturbance identification method based on S conversion and improved SVM algorithm
CN109406118A (en) * 2018-10-15 2019-03-01 华中科技大学 A kind of mechanical failure prediction method based on level convolutional neural networks
CN109632309A (en) * 2019-01-17 2019-04-16 燕山大学 Based on the rolling bearing fault intelligent diagnosing method for improving S-transformation and deep learning
CN110059357A (en) * 2019-03-19 2019-07-26 中国电力科学研究院有限公司 A kind of intelligent electric energy meter failure modes detection method and system based on autoencoder network

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105447502A (en) * 2015-11-05 2016-03-30 国家电网公司 Transient power disturbance identification method based on S conversion and improved SVM algorithm
CN109406118A (en) * 2018-10-15 2019-03-01 华中科技大学 A kind of mechanical failure prediction method based on level convolutional neural networks
CN109632309A (en) * 2019-01-17 2019-04-16 燕山大学 Based on the rolling bearing fault intelligent diagnosing method for improving S-transformation and deep learning
CN110059357A (en) * 2019-03-19 2019-07-26 中国电力科学研究院有限公司 A kind of intelligent electric energy meter failure modes detection method and system based on autoencoder network

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
时培明 等: ""基于S变换与深度学习相结合的滚动轴承故障诊断方法研究"", 《现代制造工程》 *
陈龙 等: ""基于单层SAE与SVM的滚动轴承性能退化评估"", 《机床与液压》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111832428A (en) * 2020-06-23 2020-10-27 北京科技大学 Data enhancement method applied to strip breakage fault diagnosis of cold rolling mill
CN111832428B (en) * 2020-06-23 2024-02-23 北京科技大学 Data enhancement method applied to cold rolling mill broken belt fault diagnosis
CN112304611A (en) * 2020-09-14 2021-02-02 江苏大学 Deep learning-based bearing fault diagnosis method
CN112232414A (en) * 2020-10-16 2021-01-15 广东石油化工学院 Triple concurrency fault analysis method based on X and Y dual-measurement-point spectrum data
CN112232414B (en) * 2020-10-16 2021-06-15 广东石油化工学院 Triple concurrency fault analysis method based on X and Y dual-measurement-point spectrum data
CN115358277A (en) * 2022-10-09 2022-11-18 深圳市信润富联数字科技有限公司 Bearing fault diagnosis method, device, equipment and readable storage medium

Similar Documents

Publication Publication Date Title
Saufi et al. Gearbox fault diagnosis using a deep learning model with limited data sample
Sadoughi et al. Physics-based convolutional neural network for fault diagnosis of rolling element bearings
CN111076934A (en) Method for diagnosing potential fault of bearing based on S transformation
Pan et al. LiftingNet: A novel deep learning network with layerwise feature learning from noisy mechanical data for fault classification
CN110059601B (en) Intelligent fault diagnosis method for multi-feature extraction and fusion
Xiang et al. Fault diagnosis of rolling bearing under fluctuating speed and variable load based on TCO spectrum and stacking auto-encoder
Xiong et al. Low-speed rolling bearing fault diagnosis based on EMD denoising and parameter estimate with alpha stable distribution
Guo et al. Rolling bearing fault classification based on envelope spectrum and support vector machine
CN112257530B (en) Rolling bearing fault diagnosis method based on blind signal separation and support vector machine
CN111753891B (en) Rolling bearing fault diagnosis method based on unsupervised feature learning
CN109211546B (en) Rotary machine fault diagnosis method based on noise reduction automatic encoder and increment learning
CN112304611A (en) Deep learning-based bearing fault diagnosis method
CN111238815B (en) Bearing fault identification method based on data enhancement under sample imbalance
CN114659790B (en) Identification method for bearing faults of variable-rotation-speed wind power high-speed shaft
CN112487890B (en) Bearing acoustic signal fault diagnosis method based on parallel sparse filtering
CN116793682A (en) Bearing fault diagnosis method based on iCORAL-MMD and anti-migration learning
CN113569990B (en) Strong noise interference environment-oriented performance equipment fault diagnosis model construction method
CN113639999A (en) Fault diagnosis method, system, medium and equipment for rolling bearing under variable rotating speed
Xia et al. An approach for bearing fault diagnosis based on PCA and multiple classifier fusion
CN112364706A (en) Small sample bearing fault diagnosis method based on class imbalance
Wang et al. A quantitative intelligent diagnosis method for early weak faults of aviation high-speed bearings
CN114755017B (en) Variable-speed bearing fault diagnosis method of cross-domain data driving unsupervised field shared network
CN115655717A (en) Bearing fault diagnosis method based on depth domain generalization network
CN113916535B (en) Bearing diagnosis method, system, equipment and medium based on time frequency and CNN
CN115587290A (en) Aero-engine fault diagnosis method based on variational self-coding generation countermeasure network

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
TA01 Transfer of patent application right
TA01 Transfer of patent application right

Effective date of registration: 20210207

Address after: 225400 south side of the second ring road section of Chengdong Industrial Park, Taixing City, Taizhou City, Jiangsu Province

Applicant after: JIANGSU YUANQUAN PUMP INDUSTRY Co.,Ltd.

Address before: Zhenjiang City, Jiangsu Province, 212013 Jingkou District Road No. 301

Applicant before: JIANGSU University

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

Application publication date: 20200428