CN110160781B - Test set reconstruction and prediction method for rotary machine fault classification - Google Patents
Test set reconstruction and prediction method for rotary machine fault classification Download PDFInfo
- Publication number
- CN110160781B CN110160781B CN201910525563.5A CN201910525563A CN110160781B CN 110160781 B CN110160781 B CN 110160781B CN 201910525563 A CN201910525563 A CN 201910525563A CN 110160781 B CN110160781 B CN 110160781B
- Authority
- CN
- China
- Prior art keywords
- sample
- fault
- classification
- rotary machine
- test set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
- G01M13/02—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
- 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
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
Abstract
The invention discloses a method for reconstructing and predicting a test set of rotary machine fault classification, which comprises the following steps: s01, extracting a vibration signal of a fault piece by using an acceleration sensor to acquire fault data of the rotary machine; s02, constructing index characteristics of fault classification of the rotary machine, and calculating five dimensionless indexes of waveform, pulse, margin, peak value and kurtosis of the rotary machine by using software; s03, selecting two classification algorithms of SVM and KNN for fault classification; s04, reconstructing a test set for rotary machine fault classification; and S05, establishing a prediction label determination method of the sample to be tested under the new test set. According to the method for reconstructing and predicting the test set of the fault classification of the rotary machine, provided by the invention, the test set is reconstructed by mainly utilizing the correlation among fault samples, so that the fault classification accuracy of the rotary machine can be improved.
Description
Technical Field
The invention relates to a test set reconstruction and prediction method for rotary machine fault classification, and belongs to the field of rotary machine fault and wireless sensor fault diagnosis.
Background
The Wireless Sensor Network (WSN Wireless Sensor Network) is a small-range Wireless Network formed by a plurality of distributed micro Sensor nodes with sensing, computing and communication capabilities in a self-organizing manner. The sensor nodes can sense, monitor and collect information of monitoring objects or surrounding environments in a distribution area in real time through division and cooperation among the sensor nodes, and transmit the information to a collector. The wireless sensor network node is integrated with a sensor, a data processing unit and a communication module, forms a network in a self-organizing mode, is configured with various sensors according to specific application requirements, and can measure various object information including temperature, humidity, vibration, noise, light intensity, pressure, size, speed and direction of a moving object and the like. The rotary mechanical equipment is easy to break down, and if the faults cannot be found and eliminated in time, the safety production accidents of large units are possibly caused, and great economic and property losses are caused.
Whether the running state of rotating machinery such as bearings and gears is normal or not directly affects the performance parameters such as the processing precision, the running reliability and the service life of the whole large-scale mechanical equipment. The research on the state monitoring, fault diagnosis and prediction method of the rotary machine is the basis for ensuring the safe and stable operation of the mechanical equipment, and the current diagnosis technology and signal analysis method play an important role in the real-time monitoring and fault diagnosis of the operation state of the mechanical equipment in industrial production application. With the large-scale, systematized and informationized industrial production, the requirements on the safe operation of the rotating machinery of key parts are stricter and stricter, and the method has important significance on identifying and diagnosing the faults of the gears and the bearings.
Fault diagnosis methods are generally classified into 3 types: a knowledge-based fault diagnosis method, an analytical model-based fault diagnosis method, and a signal processing-based fault diagnosis method. The knowledge-based fault diagnosis method is suitable for systems which cannot or cannot easily establish mechanism models, are insufficient in the number of sensors and lack of information. The analytical model-based method is suitable for systems with sufficient number of sensors and sufficient information, needs to fully understand the mechanism of the process, and can establish an accurate quantitative mathematical model. Knowledge-based and analytic-based fault diagnosis methods are often used to monitor situations where process parameters are relatively few. With the increase of process parameters, the current popular fault diagnosis method is a data-driven fault diagnosis method. The fault diagnosis method based on data driving analyzes and processes the process monitoring data, and completes fault diagnosis without depending on prior knowledge of a system and without knowing an accurate model of the system.
In the existing fault diagnosis method research, the method mainly focuses on the extraction of the fault data set characteristics and the improvement of a fault classification algorithm, and the method mainly reconstructs a test set by utilizing the correlation among fault samples, so that the fault classification effect is improved.
The existing fault diagnosis method research mainly focuses on the extraction of the fault data set characteristics and the improvement of a fault classification algorithm, and the fault classification effect is not ideal enough.
Disclosure of Invention
The invention aims to solve the technical problem of providing a method for reconstructing and predicting a test set by converting mechanical fault classification, which mainly utilizes the correlation among fault samples to reconstruct the test set and can improve the fault classification accuracy of rotary machines, aiming at the defects of the prior art.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows:
a method for reconstructing and predicting a test set of fault classification of a rotating machine comprises the following steps:
s01, extracting a vibration signal of a fault piece by using an acceleration sensor to acquire fault data of the rotary machine;
s02, constructing index characteristics of fault classification of the rotary machine, and calculating five dimensionless indexes of waveform, pulse, margin, peak value and kurtosis of the rotary machine by using software;
s03, selecting two classification algorithms of SVM and KNN for fault classification;
s04, reconstructing a test set for rotary machine fault classification;
and S05, establishing a prediction label determination method of the sample to be tested under the new test set.
The acquisition of rotating machine fault data adopts an EMT490 data acquisition unit.
In S02, the specific calculation method of the waveform, pulse, margin, peak and kurtosis index is as follows:
in the formula, XrmsIs the root-mean-square value of the signal,being the average amplitude of the signal, XmaxIs the maximum value of the signal, XrIs the square root amplitude of the signal, beta is the kurtosis, n denotes the number of signal samples constituting a sample, XiRepresenting the vibration amplitude, and i represents the index of the subscript of the sample point.
In S03, the fault classification structure satisfies pii>pij(j ≠ i, i, j ═ 1,2, …, n) holds, where pijRepresenting the ratio, p, of the actual type i and the classification algorithm predicted as type jiiRepresenting the proportion of the actual type i and the classification algorithm predicted to be type i, and n is the number of types of fault classification.
In S04, the specific steps include determining the correlation length l of the test sample according to the desired prediction effect, setting sample (1), sample (2), …, sample (k) as the original test set sample, and performing the reconstruction of the test sample set as follows:
wherein S is1,S2,…,SkIs a reconstructed test set sample set.
In S05, firstly, a classification algorithm is utilized to determine the prediction labels of all test samples in a sample set, a following voting mechanism is established, the number of the samples which are predicted to be of the same type in a sample set is counted, and if the statistical number of the prediction labels h in the sample set is maximum and unique, the label of a newly constructed test sample set is h; and if the predicted label is more than one with the largest statistical number, randomly selecting one label with the largest statistical number as the predicted label.
The invention has the beneficial effects that: the invention provides a test set reconstruction and prediction method for rotary machine fault classification, which reconstructs a test set by using correlation among fault samples and establishes a determination method of a prediction label of a test sample by using a voting mechanism, thereby greatly improving the prediction accuracy of the test sample.
Drawings
FIG. 1 is a flow chart of a test set reconstruction and prediction method for rotary machine fault classification according to the present invention.
Detailed Description
The present invention is further described with reference to the following examples, which are only used to more clearly illustrate the technical solutions of the present invention, but not to limit the scope of the present invention.
As shown in fig. 1, the method for reconstructing and predicting a test set of fault classification of a rotating machine provided by the present invention includes the following steps:
step one, different rolling bearings and gear box fault parts are installed on the experiment platform, vibration signals are extracted by an acceleration sensor installed on a bearing seat, and fault vibration signals are collected through an EMT490 data collector. The experimental parameters were as follows: the rotating speed is 1000r/min, the sampling frequency is 1000Hz, and the number of sampling points for calculating the dimensionless index is 1024 points.
And step two, importing the acquired vibration acceleration fault data into a computer, and reading the data by using MATLAB software. The rotary machine faults of the experiment mainly comprise five types: the vibration signals under the five states are respectively sampled by the tooth lack of the big gear, the tooth lack of the small gear, the abrasion of the bearing inner ring, the abrasion of the bearing outer ring and the normal state, and the number of sampling points in each state is 102400.
Step three, calculating 5 dimensionless indexes by using 1024 continuous sampling points: a waveform index, a peak index, a pulse index, a margin index and a kurtosis index are adopted, and the decision attribute value of the rotary mechanical running state mark sample in the second implementation mode step is adopted;
in the formula, XrmsIs the root-mean-square value of the signal,being the average amplitude of the signal, XmaxIs the maximum value of the signal, XrIs the square root amplitude of the signal, beta is the kurtosis, n denotes the number of signal samples constituting a sample, XiRepresenting the vibration amplitude, and i represents the index of the subscript of the sample point.
Selecting two classification algorithms of SVM and KNN to classify and predict the faults, wherein 40% of samples of each fault type are randomly selected as a training set, and the rest 60% of samples are used as test samples to perform classification predictionAnd (6) measuring. Fault classification structure satisfies pii>pij(j ≠ i, i, j ═ 1,2, …, n) holds, where pijRepresenting the ratio, p, of the actual type i and the classification algorithm predicted as type jiiRepresenting the proportion of the actual type i and the classification algorithm predicted to be type i, and n is the number of types of fault classification.
Step five, reconstructing the test set according to the step three of the invention contents, specifically, determining the correlation length l of the test sample according to the prediction effect to be achieved, setting sample (1), sample (2), …, and sample (k) as the original test set sample, and reconstructing the test sample set according to the following modes:
wherein S is1,S2,…,SkIs a reconstructed test set sample set.
The number of original test samples is 60 for each fault type, when l is 10, the number of reconstructed test set samples is 51, when l is 20, the number of reconstructed test set samples is 41, and when l is 30, the number of reconstructed test set samples is 31.
And step six, according to the method for determining the prediction label of the sample to be tested under the new test set given in the step five of the invention content, the prediction accuracy of the original test set and the reconstructed test set is counted. Firstly, determining a prediction label of each test sample in a sample set by using a classification algorithm, establishing a voting mechanism, counting the number of the prediction labels h in the sample set into the same type, and if the statistical number of the prediction labels h in the sample set is maximum and unique, setting the label of a newly constructed test sample set as h; if the statistical number of the predicted labels is more than one, randomly selecting one label with the largest statistical number as the predicted label, and comparing the classification accuracy rates, wherein the results are shown in table 1.
TABLE 1
As can be seen from table 1, with the KNN algorithm and the support vector machine, the prediction accuracy of the fault reconstruction test set of the rotary machine is higher than that of the original test set, and the classification accuracy increases with the increase of the correlation length.
The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.
Claims (3)
1. A method for reconstructing and predicting a test set of fault classification of a rotating machine is characterized by comprising the following steps: the method comprises the following steps:
s01, extracting a vibration signal of a fault piece by using an acceleration sensor to acquire fault data of the rotary machine;
s02, constructing index characteristics of fault classification of the rotary machine, and calculating five dimensionless indexes of waveform, pulse, margin, peak value and kurtosis of the rotary machine by using software;
s03, selecting two classification algorithms of SVM and KNN for fault classification, wherein the fault classification structure satisfies pii>pijWhere j ≠ i, i, j ≠ 1,2, …, n, where pijRepresenting the ratio, p, of the actual type i and the classification algorithm predicted as type jiiRepresenting the proportion that the actual type is i and the classification algorithm is predicted to be type i, n is the number of the types of fault classification, if the fault classification structure does not satisfy pii>pijIf j ≠ i, i, j ≠ 1,2, …, n, returning to S02 to reconstruct index features of fault classification of the rotary machine;
s04, reconstructing a test set for rotary machine fault classification;
s05, establishing a new method for determining the predicted label of the sample to be tested under the test set, firstly, determining the predicted label of each test sample in the sample set by using a classification algorithm, establishing a following voting mechanism, counting the number of the predicted labels h in the sample set into the same type, and if the statistical number of the predicted labels h in the sample set is the maximum and unique number, setting the label of the newly constructed test sample set as h; and if the statistical number of the predicted labels is more than one, randomly selecting one label with the largest statistical number as the predicted label of each test sample in the sample set.
2. The method of claim 1, wherein the method comprises: in S02, the specific calculation method of the waveform, pulse, margin, peak and kurtosis index is as follows:
in the formula, XrmsIs the root-mean-square value of the signal,being the average amplitude of the signal, XmaxIs the maximum value of the signal, XrIs the square root amplitude of the signal, beta is the kurtosis, n denotes the number of signal samples constituting a sample, XiRepresenting the vibration amplitude, and i represents the index of the subscript of the sample point.
3. The method of claim 1, wherein the method comprises: in S04, the specific steps include determining the correlation length l of the test sample according to the desired prediction effect, setting sample (1), sample (2), …, sample (m) as the original test set sample, and performing the reconstruction of the test sample set as follows:
wherein S is1,S2,…,SmIs a reconstructed test set sample set.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910525563.5A CN110160781B (en) | 2019-06-18 | 2019-06-18 | Test set reconstruction and prediction method for rotary machine fault classification |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910525563.5A CN110160781B (en) | 2019-06-18 | 2019-06-18 | Test set reconstruction and prediction method for rotary machine fault classification |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110160781A CN110160781A (en) | 2019-08-23 |
CN110160781B true CN110160781B (en) | 2020-12-04 |
Family
ID=67625801
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910525563.5A Active CN110160781B (en) | 2019-06-18 | 2019-06-18 | Test set reconstruction and prediction method for rotary machine fault classification |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110160781B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112183344B (en) * | 2020-09-28 | 2021-06-01 | 广东石油化工学院 | Large unit friction fault analysis method and system based on waveform and dimensionless learning |
CN112270227A (en) * | 2020-10-16 | 2021-01-26 | 广东石油化工学院 | Oil film whirl and friction concurrent fault analysis method and analysis system |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106371427B (en) * | 2016-10-28 | 2019-03-29 | 浙江大学 | Industrial process Fault Classification based on analytic hierarchy process (AHP) and fuzzy Fusion |
CN106646096B (en) * | 2016-11-15 | 2019-07-02 | 国网四川省电力公司广安供电公司 | Transformer fault classification and recognition methods based on vibration analysis method |
CN108073158A (en) * | 2017-12-05 | 2018-05-25 | 上海电机学院 | Based on PCA and KNN density algorithm Wind turbines Method for Bearing Fault Diagnosis |
CN109323860A (en) * | 2018-10-31 | 2019-02-12 | 广东石油化工学院 | A kind of rotating machinery gearbox fault data set optimization method |
CN109489977B (en) * | 2018-12-28 | 2021-03-05 | 西安工程大学 | KNN-AdaBoost-based bearing fault diagnosis method |
-
2019
- 2019-06-18 CN CN201910525563.5A patent/CN110160781B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN110160781A (en) | 2019-08-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108681747B (en) | Rotary machine fault diagnosis and state monitoring system and method based on deep learning | |
CN111353482B (en) | LSTM-based fatigue factor recessive anomaly detection and fault diagnosis method | |
CN111623868B (en) | Convolutional neural network construction method for rail corrugation identification | |
CN110287552B (en) | Motor bearing fault diagnosis method and system based on improved random forest algorithm | |
CN111637964B (en) | Rail corrugation identification method | |
CN110110768B (en) | Rolling bearing fault diagnosis method based on parallel feature learning and multiple classifiers | |
CN110333077B (en) | Rolling bearing fault diagnosis method based on approximate support vector | |
CN111562108A (en) | Rolling bearing intelligent fault diagnosis method based on CNN and FCMC | |
CN102155988B (en) | Equipment monitoring and diagnosing method | |
CN110008898B (en) | Industrial equipment data edge processing method based on symbol and convolutional neural network | |
CN109946080B (en) | Mechanical equipment health state identification method based on embedded circulation network | |
CN113008559B (en) | Bearing fault diagnosis method and system based on sparse self-encoder and Softmax | |
CN113076834B (en) | Rotating machine fault information processing method, processing system, processing terminal, and medium | |
CN110160781B (en) | Test set reconstruction and prediction method for rotary machine fault classification | |
CN112364706A (en) | Small sample bearing fault diagnosis method based on class imbalance | |
CN114705432B (en) | Method and system for evaluating health state of explosion-proof motor bearing | |
CN114462480A (en) | Multi-source sensor rolling mill fault diagnosis method based on non-equilibrium data set | |
CN114330096B (en) | Rail transit motor bearing fault degree classification model construction method based on deep learning | |
CN116538092B (en) | Compressor on-line monitoring and diagnosing method, device, equipment and storage medium | |
CN110056640A (en) | Speed reducer wireless malfunction diagnostic method based on acceleration signal and edge calculations | |
CN116956215A (en) | Fault diagnosis method and system for transmission system | |
CN117036732A (en) | Electromechanical equipment detection system, method and equipment based on fusion model | |
CN116644304A (en) | Similarity-based truck bearing abnormal feature extraction and classification method | |
CN113807431A (en) | Intelligent spindle state evaluation method and system based on multi-source information fusion | |
CN111595582A (en) | Fault diagnosis method for rolling bearing |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |