CN111397896B - Fault diagnosis method and system for rotary machine and storage medium - Google Patents

Fault diagnosis method and system for rotary machine and storage medium Download PDF

Info

Publication number
CN111397896B
CN111397896B CN202010154567.XA CN202010154567A CN111397896B CN 111397896 B CN111397896 B CN 111397896B CN 202010154567 A CN202010154567 A CN 202010154567A CN 111397896 B CN111397896 B CN 111397896B
Authority
CN
China
Prior art keywords
fault
fault diagnosis
signals
sub
vector
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
Application number
CN202010154567.XA
Other languages
Chinese (zh)
Other versions
CN111397896A (en
Inventor
刘颉
曹贯男
周凯波
潘浩
张凯锋
葛子月
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.)
Huazhong University of Science and Technology
Original Assignee
Huazhong University of Science and Technology
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 Huazhong University of Science and Technology filed Critical Huazhong University of Science and Technology
Priority to CN202010154567.XA priority Critical patent/CN111397896B/en
Publication of CN111397896A publication Critical patent/CN111397896A/en
Application granted granted Critical
Publication of CN111397896B publication Critical patent/CN111397896B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G01M13/021—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
    • 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
    • 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)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Acoustics & Sound (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

The invention discloses a fault diagnosis method, a fault diagnosis system and a storage medium for a rotary machine, wherein the method comprises the following steps: acquiring vibration signals of a rotary machine in a normal state and a fault state, and dividing the vibration signals into a training sample set and a sample to be tested; decomposing the obtained vibration signal by using a wavelet packet transformation method to obtain a series of sub-signals with different frequency bands; calculating the symbolic dynamics entropy value of the sub-signal to obtain a fault feature vector; taking the fault feature vector of the training sample set as input, taking the fault type label of the training sample set as output, and training to obtain a fault diagnosis model based on a LightGBM classifier model; and inputting the fault characteristic vector of the sample to be detected into the fault diagnosis model, thereby obtaining a fault diagnosis result of the sample to be detected. According to the method, the wavelet packet decomposition is combined with the symbolic dynamics entropy to effectively extract fault characteristics, and then the LightGBM classifier model is used for fault identification and classification, so that the calculation efficiency and the classification accuracy are improved.

Description

Fault diagnosis method and system for rotary machine and storage medium
Technical Field
The invention belongs to the technical field of fault diagnosis of rotary machines, and particularly relates to a fault diagnosis method and system of a rotary machine and a storage medium.
Background
Rotary machines have been widely used in modern manufacturing and industrial processes as a key component of transmission systems. In most practical applications, rotary machines operate under harsh or complex conditions, such as high temperature, high pressure environments, variable speeds and variable loads. Long runs can result in various damage and failures that can affect system performance and can severely damage the machine.
Rotary machine fault diagnosis methods can be divided into two types: feature-based methods and feature-based learning methods. (1) The feature-based method is to select and calculate fault features according to prior knowledge and engineering experience, and then input the fault features into a classification algorithm for fault diagnosis, and comprises four basic steps: signal preprocessing, feature calculation, feature selection and fault classification, it can be seen that the method is mainly based on signal processing, such as fourier transform and wavelet transform; however, this method is limited in practical application by a priori knowledge and experience due to the non-linearity and non-stationarity of the vibration signal; furthermore, some of these methods require the application of multiple signal preprocessing methods and the calculation of multiple statistical signatures to obtain sufficient fault signatures, which also increases the complexity of the method implementation. (2) The method based on feature learning mainly converts a one-dimensional original signal into a two-dimensional image through a dimension conversion method, then automatically performs feature learning on the two-dimensional image through a deep learning method such as a multi-scale network and a convolutional neural network, extracts fault information, and finally performs classification through a classifier; however, the method based on feature learning requires a large number of samples and a large amount of time for model training, and is inefficient.
Disclosure of Invention
Aiming at the defects and improvement requirements of the prior art, the invention provides a fault diagnosis method for a rotary machine, and aims to solve the technical problems that fault feature extraction is complex to implement, the calculation efficiency is low, the model training time is long, and the method is difficult to adapt to complex working conditions in the prior art.
To achieve the above object, according to one aspect of the present invention, there is provided a rotary machine fault diagnosis method including the steps of:
s1: acquiring vibration signals of a rotary machine in a normal state and a fault state, and dividing the vibration signals into a training sample set and a sample to be tested;
s2: decomposing the vibration signal obtained in the step S1 by using a wavelet packet transformation method to obtain a series of sub-signals with different frequency bands;
s3: calculating the symbolic dynamics entropy value of the sub-signal to obtain a fault feature vector;
s4: taking the fault feature vector of the training sample set as the input of a LightGBM classifier model, taking the fault type label of the training sample set as the output of the LightGBM classifier model, and training to obtain a fault diagnosis model;
s5: and inputting the fault characteristic vector of the sample to be detected into the fault diagnosis model, thereby obtaining a fault diagnosis result of the sample to be detected.
Further, the method of step S2 includes:
given a scale function phi (n) and a wavelet function psi (n), let
Figure BDA0002403617630000021
Wavelet packet
Figure BDA0002403617630000022
Is defined as:
Figure BDA0002403617630000023
wherein i is 0,1,2 … …, n represents the value of a time sequence corresponding to a certain time, k represents time or position parameter, Z is integer set, hkDenotes a low-pass filter, gkDenotes a high-pass filter, hkAnd gkIs a pair of conjugate filters and satisfies gk＝(-1)kh1-k；
The vibration signal x (n) is decomposed into sub-signals using wavelet packets
Figure BDA0002403617630000031
Wherein j represents the number of decomposition layers, p represents the number of signals and p is 0,1, …,2j-1，
Figure BDA0002403617630000032
And represents the p-th wavelet packet coefficient of the j-th layer.
Further, the method of step S3 includes:
by passingThe normal cumulative distribution function maps the sub-signals to yiWherein i is 1,2, …, N and 0 < yi＜1；
The symbol time sequence corresponding to the sub-signal is si＝round(C·yi+0.5), where round (·) denotes an integer function, C is the number of symbols;
constructing embedding vectors for the time series of symbols, each embedding vector being represented as: vi m,λ＝{si,si+λ,…,si+(m-1)λWhere m is the embedding dimension and λ is the time delay;
the embedding vector Vi m,λAnd status mode
Figure BDA0002403617630000033
One-to-one correspondence, where ξ represents the state of each element embedded within a vector, and ξ represents the state of each element within an embedded vector1＝si,ξ2＝si+λ,…,ξm＝si+(m-1)λThe probability of each state pattern is:
Figure BDA0002403617630000034
where | represents the number of elements in the set;
the obtained fault feature vector is:
Figure BDA0002403617630000035
further, in step S1, the vibration signal of the rotating machine is collected by an accelerometer.
Further, the fault type label in the step S4 is used to distinguish the fault type and the fault severity.
In another aspect, the present invention provides a fault diagnosis system for a rotary machine, including:
the signal acquisition unit is used for acquiring vibration signals of the rotary machine in a normal state and a fault state and dividing the vibration signals into a training sample set and a sample to be tested;
the signal decomposition unit is used for decomposing the acquired vibration signals by using a wavelet packet transformation method to obtain a series of sub-signals with different frequency bands;
the characteristic vector calculating unit is used for calculating the symbolic dynamics entropy value of the sub-signals to obtain fault characteristic vectors;
the model training unit is used for taking the fault feature vector of the training sample set as the input of the LightGBM classifier model, taking the fault type label of the training sample set as the output of the LightGBM classifier model, and training to obtain a fault diagnosis model;
and the fault diagnosis unit is used for inputting the fault characteristic vector of the sample to be detected into the fault diagnosis model so as to obtain a fault diagnosis result of the sample to be detected.
Another aspect of the present invention provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the rotating machine fault diagnosis method as described above.
Generally, by the above technical solution conceived by the present invention, the following beneficial effects can be obtained:
(1) according to the method, an original vibration signal is decomposed into a series of sub-signals with different frequency bands through a wavelet packet, symbol dynamics entropy values of the sub-signals are calculated to obtain a fault characteristic vector, and meanwhile, a fault diagnosis model based on a LightGBM classifier model is trained, so that the fault type and the fault degree are effectively identified.
(2) According to the invention, an original signal is decomposed into a low-frequency part and a high-frequency part through wavelet packet decomposition, and then the wavelet packet decomposition is sequentially used for the low-frequency part and the high-frequency part, so that different frequency information is obtained according to different decomposition layer numbers; the symbolic dynamics entropy is a method for evaluating dynamic characteristics of a time sequence, has higher calculation efficiency, and can extract high-quality fault features; the combination of the two can effectively extract fault characteristics, and is particularly suitable for extracting the fault characteristics under complex working conditions.
(3) Compared with a plurality of decomposition methods or a plurality of characteristic methods, the rotary machine fault diagnosis method provided by the invention has the advantages of easiness in implementation and capability of processing big data fault diagnosis.
Drawings
FIG. 1 is a flow chart of a method of fault diagnosis for a rotating machine provided by the present invention;
FIG. 2 is a diagram of a fault diagnosis multi-level confusion matrix provided by an embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating classification accuracy of a fault diagnosis method for a rotating machine according to an embodiment of the present invention under a complex working condition data set;
FIG. 4 is a diagram illustrating classification accuracy comparing different signal decomposition methods according to an embodiment of the present invention;
FIG. 5 is a graph illustrating classification accuracy comparing different entropy-based methods provided by embodiments of the present invention;
fig. 6 is a diagram illustrating classification accuracy comparing different classifier methods according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
On one hand, the invention provides a rotary machine fault diagnosis method, which effectively extracts fault characteristics by combining wavelet packet decomposition with symbolic dynamics entropy, and then identifies and classifies faults by using a LightGBM classifier model, thereby improving the calculation efficiency and the classification accuracy.
Fig. 1 is a flowchart of a fault diagnosis method for a rotary machine according to the present invention, which specifically includes the following steps:
s1: acquiring vibration signals of a rotary machine in a normal state and a fault state, and dividing the vibration signals into a training sample set and a sample to be tested;
wherein the fault states include an inner ring fault, an outer ring fault and a ball fault;
specifically, vibration signals of the rotating machine in a normal state and a fault state are collected through an accelerometer, and sample points containing the vibration signals in the normal state and the fault state are obtained; randomly selecting part of sample points to form a training sample set, forming a sample to be tested by the rest sample points, and setting the ratio of the training sample set to the test sample to be 2.
S2: decomposing the vibration signal obtained in the step S1 by using a wavelet packet transformation method to obtain a series of sub-signals with different frequency bands;
specifically, given a scale function φ (n) and a wavelet function ψ (n), let
Figure BDA0002403617630000061
Wavelet packet
Figure BDA0002403617630000062
Is defined as:
Figure BDA0002403617630000063
wherein i is 0,1,2 … …, n represents the value of a time sequence corresponding to a certain time, k represents time or position parameter, Z is integer set, hkDenotes a low-pass filter, gkDenotes a high-pass filter, hkAnd gkIs a pair of conjugate filters and satisfies gk＝(-1)kh1-k；
The amplitude-frequency characteristics of the high-pass filter and the low-pass filter are symmetric about pi/2 as an axis, like a quadrature mirror symmetric filter bank (QMF), but the phase-frequency characteristics are more than the QMF by a conjugate relationship, and are therefore called a conjugate filter bank.
The vibration signal x (n) is decomposed into sub-signals using wavelet packets
Figure BDA0002403617630000064
Wherein j represents the number of decomposition layers, p represents the number of signals and p is 0,1, …,2j-1，
Figure BDA0002403617630000065
And represents the p-th wavelet packet coefficient of the j-th layer.
S3: calculating the symbolic dynamics entropy value of the sub-signal to obtain a fault feature vector;
in particular, the sub-signals are mapped to y by a normal cumulative distribution functioniWherein i is 1,2, …, N and 0 < yi＜1；
The symbol time sequence corresponding to the sub-signal is si＝round(C·yi+0.5), where round (·) denotes an integer function, C is the number of symbols;
constructing embedding vectors for the time series of symbols, each embedding vector being represented as: vi m,λ＝{si,si+λ,…,si+(m-1)λWhere m is the embedding dimension and λ is the time delay;
the embedding vector Vi m,λAnd status mode
Figure BDA0002403617630000071
One-to-one correspondence, where ξ represents the state of each element embedded within a vector, and ξ represents the state of each element within an embedded vector1＝si,ξ2＝si+λ,…,ξm＝si+(m-1)λThe probability of each state pattern is:
Figure BDA0002403617630000072
where | represents the number of elements in the set;
the obtained fault feature vector is:
Figure BDA0002403617630000073
it should be noted that, according to the embedding dimension and the time delay, the symbol sequence can be divided into a series of sub-vectors, wherein the symbol arrangement mode of each sub-vector is unique, and each arrangement is a state mode;
the fault feature vector may also be normalized: NSDE (x, m, C, λ) ═ SDE (x, m, C, λ)/logCm，0≤NSDE(x,m,C,λ)≤1。
S4: taking the fault feature vector of the training sample set as the input of a LightGBM classifier model, taking the fault type label of the training sample set as the output of the LightGBM classifier model, and training to obtain a fault diagnosis model;
LightGBM model parameters were set, the main parameters are shown in table 1.
TABLE 1
Figure BDA0002403617630000074
S5: and inputting the fault characteristic vector of the sample to be detected into the fault diagnosis model, thereby obtaining a fault diagnosis result of the sample to be detected.
Specifically, the sample to be tested also needs to be preprocessed, the original vibration signals under different working conditions are processed in steps S1-S3 to obtain fault feature vectors, and the fault feature vectors are substituted into the fault diagnosis model trained in step S4 to obtain corresponding fault diagnosis results.
The utility of the present invention is further verified below.
Example 1: validity verification
In order to verify the effectiveness of the fault diagnosis method provided by the present invention, the present embodiment uses a bearing data set of a bearing data center of the University of caes Western University (CWRU) of Case Western Reserve. The fault types of the rolling bearing are normal, inner ring defect, outer ring defect and ball defect, wherein the fault severity is simulated by electric spark machining, and the fault diameters are 7, 14 and 21(mil) respectively. And acquiring a vibration signal at a 6 o' clock position at the end of the driving motor, wherein the sampling frequency is 12 kHz. The data selected by the experiment are 4 working conditions in total, each working condition has 10 bearing states, and the detailed conditions of the bearing states are shown in table 2.
TABLE 2
Figure BDA0002403617630000081
The method comprises the following specific steps:
(1) data acquisition and wavelet packet decomposition
Each bearing condition had 48 data samples per condition, and the total data set used for the experiment was 1920 data samples, with each data sample having a length of 1500 data points. 1280 was randomly selected as a training sample and 640 additional data samples were selected as test samples in the experiment. And (3) carrying out wavelet packet decomposition on the experimental data set, setting the number of layers of the wavelet packet decomposition to be 5, and decomposing the original vibration signal to obtain a series of sub-signals.
(2) Computing the sign-dynamics entropy of subsignals
Calculating sign dynamics entropy for the decomposed subsignals, wherein three parameters for calculating sign dynamics entropy are set as: the symbol number C is 10, the embedding dimension m is 3, and the time delay λ is 1. The sign dynamics entropy can effectively extract dynamic characteristics of the time sequence, reflects the complexity of the time sequence, has higher calculation efficiency than other entropy measures, and obtains fault characteristic vectors by calculating the sign dynamics entropy of the sub-signals.
(3) Diagnostic model training
And (3) taking the characteristic vector obtained in the step (2) as an input, substituting the characteristic vector into a LightGBM classifier model, and performing model parameterization initial setting and training to obtain an optimal diagnosis model.
(4) State recognition
And (3) processing the original vibration signal in the steps (1) and (2) to obtain a fault characteristic vector, substituting the fault characteristic vector into the optimal diagnosis model trained in the step (3) to obtain a corresponding fault diagnosis result, wherein a fault diagnosis multilayer confusion matrix diagram is shown in fig. 2. Fig. 2 shows that the fault diagnosis method provided by the present invention has a classification accuracy of 100% for the bearing state types 1,2, 3, 4, 5, 7, 8, and 10, a classification accuracy of 97% for the bearing state type 6, and a 3% probability of misdiagnosis as the bearing state type 9, and a classification accuracy of 97% for the bearing state type 9, and a 3% probability of misdiagnosis as the bearing state type 6, and has a better diagnosis performance.
Example 2: verification of effectiveness under complex working conditions
In order to further highlight the performance of the proposed method, the present invention is studied from four aspects for the complex behavior of the actual data set: additional verification is performed on the actual application data set, and the actual application data set is used for comparing different decomposition methods, entropy-based methods and classification algorithms.
(1) Verification using actual data set
Since the four operating conditions of the CWRU bearing dataset are similar in speed, the actual dataset with more complex operating conditions was used to further verify the validity of the proposed method of the present invention. The experimental platform consists of an electromagnetic brake, a torque sensor, a single-stage reducer, a brake controller and a servo motor. Gears with different crack lengths (including 0, 5, 10, 15 mm) were used and the sampling frequency was 5 kHz. In the invention, ten data sets consisting of data from twenty working conditions are adopted for experimental verification, and the detailed information of the twenty working conditions and the ten data sets is respectively shown in tables 3 and 4.
TABLE 3
Figure BDA0002403617630000101
TABLE 4
Figure BDA0002403617630000102
The data sets from A1 to A4 correspond to single speed and multiple load conditions. In comparison, the data sets from A5 to A9 correspond to a single load and multiple speeds, and the data set A10 corresponds to the most complex operating conditions, i.e., multiple speeds and multiple loads. The next three comparative experiments will be compared using data set a 10. Each sample was 1500 data points in length, with no overlap, for a total of 3200 samples.
The classification results of the experiments are shown in fig. 3, and the results show that the training accuracy of ten data sets reaches 100%, the test accuracy from A1 to A4 is higher than 94.50%, and the test accuracy from A5 to A9 is higher than 96.56%. Under the most complicated condition A10, the test accuracy was 96.66%. The highest test accuracy of A5 is up to 99.69%. The F1 score of the test set is consistent with the test accuracy, and the effectiveness of the method is verified in the actual data set of the single-stage gearbox.
(2) Comparison with different signal decomposition methods
The validity of Wavelet Packet Decomposition (WPD) has been verified in example 1, and in order to highlight its advantages, three decomposition methods are used: empirical Mode Decomposition (EMD), Local Mean Decomposition (LMD), and Variational Mode Decomposition (VMD) were compared. The comparison results are shown in fig. 4 and table 5.
TABLE 5
Figure BDA0002403617630000111
Experimental results show that the WPD testing accuracy is the highest and reaches 96.66%, and the WPD testing accuracy is at least 6.66% higher than that of the other three methods. The feature extraction time of EMD is minimum, only 0.01 s/sample, the time required by VMD is maximum (0.85 s/sample), and the WPD training time is moderate (0.17 s/sample). The standard deviation of the test accuracy using the WPD method was the second smallest (1.07%). Therefore, the WPD method can obtain the best classification accuracy with the total time consumption at a moderate level, compared to the other three methods.
(3) Comparison with different entropy-based methods
Entropy-based fault feature extraction methods have received wide attention in the field of mechanical fault diagnosis, such as Sample Entropy (SE), Fuzzy Entropy (FE), and Permutation Entropy (PE). In the invention, SE, FE and PE of each sub-signal of the original vibration signal after being decomposed by WPD are respectively calculated, and the classification result of the SE, FE and PE is compared with the Symbol Dynamic Entropy (SDE) of the proposed method. Where the parameters of SE and FE are set to embedding dimension m 2, time delay λ 1, and the embedding dimension and time delay of PE are set to m 5 and λ 1, respectively. The comparison results are shown in fig. 5 and table 6.
TABLE 6
Figure BDA0002403617630000121
The experimental result shows that the test accuracy of the method (WPD-SDE) provided by the invention is highest and reaches 96.66%. As can be seen from Table 6, the corresponding test accuracy and the standard deviation of the training time using WPD-FE are the smallest, but the time consumption of feature extraction by WPD-FE is 1.89 s/sample, which is 11 times that of the method provided by the present invention. In summary, among these entropy-based methods, the method proposed by the present invention has the least feature extraction time consumption, the highest classification accuracy and smaller standard deviation of test accuracy.
(4) Comparison with different classifier methods
The state recognition is an important step in fault diagnosis, and the classification algorithm can directly influence the final diagnosis result. In the invention, the LightGBM algorithm is compared with a Support Vector Machine (SVM), a Random Forest (RF) and an extreme gradient boosting (XGBoost) algorithm. The main hyperparameters of SVM, RF and XGBoost were optimized using grid search, with the hyperparameter settings as shown in table 7.
TABLE 7
Figure BDA0002403617630000122
Figure BDA0002403617630000131
Fig. 6 illustrates the comparison of different classification methods, using LightGBM as classifier, the highest test accuracy, 96.66%, can be achieved. Table 8 lists the training times, test times and standard deviations of test accuracy for the different classifiers. Among these classifiers, the training time of the LightGBM model is the shortest, only 0.64 s. The test time difference is larger compared to the training time, the minimum is LightGBM, which is 3.13 ms. Accordingly, the standard deviation of the test accuracy for all methods is small, ranging from 0.82% to 1.12%. It can also be seen that the influence of the classification method on the state recognition is not as important as the feature extraction method, which further proves the effectiveness of the feature extraction method provided by the invention.
TABLE 8
Figure BDA0002403617630000132
According to the invention, the high-quality fault characteristics are extracted from the non-stationary signals according to the characteristics of wavelet packet decomposition and symbolic dynamic entropy, and the problem of complex working conditions in fault diagnosis of the rotary machine is effectively solved. The method provided by the invention is a rotary machine fault diagnosis method based on wavelet packet decomposition, symbolic dynamic entropy and LightGBM, takes fault diagnosis of a rolling bearing and a gear as example verification, can be popularized to general rotary machines, machining and manufacturing, equipment maintenance and the like in the practical application process, and has good engineering practicability.
Another aspect of an embodiment of the present invention provides a fault diagnosis system for a rotary machine, including:
the signal acquisition unit is used for acquiring vibration signals of the rotary machine in a normal state and a fault state and dividing the vibration signals into a training sample set and a sample to be tested;
the signal decomposition unit is used for decomposing the acquired vibration signals by using a wavelet packet transformation method to obtain a series of sub-signals with different frequency bands;
the characteristic vector calculating unit is used for calculating the symbolic dynamics entropy value of the sub-signals to obtain fault characteristic vectors;
the model training unit is used for taking the fault feature vector of the training sample set as the input of the LightGBM classifier model, taking the fault type label of the training sample set as the output of the LightGBM classifier model, and training to obtain a fault diagnosis model;
and the fault diagnosis unit is used for inputting the fault characteristic vector of the sample to be detected into the fault diagnosis model so as to obtain a fault diagnosis result of the sample to be detected.
Another aspect of the present invention provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the rotating machine fault diagnosis method as described above.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims (5)

1. A method of fault diagnosis for a rotating machine, comprising the steps of:
s1: acquiring vibration signals of a rotary machine in a normal state and a fault state, and dividing the vibration signals into a training sample set and a sample to be tested;
s2: decomposing the vibration signal obtained in the step S1 by using a wavelet packet transformation method to obtain a series of sub-signals with different frequency bands; the method specifically comprises the following steps:
given a scale function phi (n) and a wavelet function psi (n), let
Figure FDA0002756186460000014
Wavelet packet
Figure FDA0002756186460000015
Is defined as:
Figure FDA0002756186460000011
wherein i is 0,1,2 … …, n represents the value of a time sequence corresponding to a certain time, k represents time or position parameter, Z is integer set, hkDenotes a low-pass filter, gkDenotes a high-pass filter, hkAnd gkIs a pair of conjugate filters and satisfiesgk＝(-1)kh1-k；
The vibration signal x (n) is decomposed into sub-signals using wavelet packets
Figure FDA0002756186460000012
Wherein j represents the number of decomposition layers, p represents the number of signals and p is 0,1, …,2j-1，
Figure FDA0002756186460000013
Representing the p-th wavelet packet coefficient of the j-th layer;
s3: calculating the symbolic dynamics entropy value of the sub-signal to obtain a fault feature vector; the method specifically comprises the following steps:
mapping the sub-signals to y by a normal cumulative distribution functioniWherein i is 1,2, …, N and 0 < yi＜1；
The symbol time sequence corresponding to the sub-signal is si＝round(C·yi+0.5), where round (·) denotes an integer function, C is the number of symbols;
constructing embedding vectors for the time series of symbols, each embedding vector being represented as: vi m,λ＝{si,si+λ,…,si+(m-1)λWhere m is the embedding dimension and λ is the time delay;
the embedding vector Vi m,λAnd status mode
Figure FDA0002756186460000021
One-to-one correspondence, where ξ represents the state of each element embedded within a vector, and ξ represents the state of each element within an embedded vector1＝si,ξ2＝si+λ,…,ξm＝si+(m-1)λThe probability of each state pattern is:
Figure FDA0002756186460000022
where | represents the number of elements in the set;
the obtained fault feature vector is:
Figure FDA0002756186460000023
s4: taking the fault feature vector of the training sample set as the input of a LightGBM classifier model, taking the fault type label of the training sample set as the output of the LightGBM classifier model, and training to obtain a fault diagnosis model;
s5: and inputting the fault characteristic vector of the sample to be detected into the fault diagnosis model, thereby obtaining a fault diagnosis result of the sample to be detected.
2. The fault diagnosis method according to claim 1, wherein the vibration signal of the rotating machine in step S1 is collected by an accelerometer.
3. The fault diagnosing method as claimed in claim 1, wherein the fault type label in step S4 is used to distinguish the fault type and the fault severity.
4. A rotary machine fault diagnosis system, characterized by comprising the following units:
the signal acquisition unit is used for acquiring vibration signals of the rotary machine in a normal state and a fault state and dividing the vibration signals into a training sample set and a sample to be tested;
the signal decomposition unit is used for decomposing the acquired vibration signals by using a wavelet packet transformation method to obtain a series of sub-signals with different frequency bands; in particular for the use in the manufacture of,
given a scale function phi (n) and a wavelet function psi (n), let
Figure FDA0002756186460000024
Wavelet packet
Figure FDA0002756186460000025
Is defined as:
Figure FDA0002756186460000031
wherein i is 0,1,2 … …, n represents the value of a time sequence corresponding to a certain time, k represents time or position parameter, Z is integer set, hkDenotes a low-pass filter, gkDenotes a high-pass filter, hkAnd gkIs a pair of conjugate filters and satisfies gk＝(-1)kh1-k；
The vibration signal x (n) is decomposed into sub-signals using wavelet packets
Figure FDA0002756186460000032
Wherein j represents the number of decomposition layers, p represents the number of signals and p is 0,1, …,2j-1，
Figure FDA0002756186460000036
Representing the p-th wavelet packet coefficient of the j-th layer;
the characteristic vector calculating unit is used for calculating the symbolic dynamics entropy value of the sub-signals to obtain fault characteristic vectors; in particular for the use in the manufacture of,
mapping the sub-signals to y by a normal cumulative distribution functioniWherein i is 1,2, …, N and 0 < yi＜1；
The symbol time sequence corresponding to the sub-signal is si＝round(C·yi+0.5), where round (·) denotes an integer function, C is the number of symbols;
constructing embedding vectors for the time series of symbols, each embedding vector being represented as: vi m,λ＝{si,si+λ,…,si+(m-1)λWhere m is the embedding dimension and λ is the time delay;
the embedding vector Vi m,λAnd status mode
Figure FDA0002756186460000033
One-to-one correspondence, where ξ represents the state of each element embedded within a vector, and ξ represents the state of each element within an embedded vector1＝si,ξ2＝si+λ,…,ξm＝si+(m-1)λEach ofThe probability of an individual state pattern is:
Figure FDA0002756186460000034
where | represents the number of elements in the set;
the obtained fault feature vector is:
Figure FDA0002756186460000035
the model training unit is used for taking the fault feature vector of the training sample set as the input of the LightGBM classifier model, taking the fault type label of the training sample set as the output of the LightGBM classifier model, and training to obtain a fault diagnosis model;
and the fault diagnosis unit is used for inputting the fault characteristic vector of the sample to be detected into the fault diagnosis model so as to obtain a fault diagnosis result of the sample to be detected.
5. A computer-readable storage medium, characterized in that a computer program is stored thereon, which, when being executed by a processor, implements the fault diagnosis method according to any one of claims 1 to 3.
CN202010154567.XA 2020-03-08 2020-03-08 Fault diagnosis method and system for rotary machine and storage medium Active CN111397896B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010154567.XA CN111397896B (en) 2020-03-08 2020-03-08 Fault diagnosis method and system for rotary machine and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010154567.XA CN111397896B (en) 2020-03-08 2020-03-08 Fault diagnosis method and system for rotary machine and storage medium

Publications (2)

Publication Number Publication Date
CN111397896A CN111397896A (en) 2020-07-10
CN111397896B true CN111397896B (en) 2020-12-18

Family

ID=71434117

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010154567.XA Active CN111397896B (en) 2020-03-08 2020-03-08 Fault diagnosis method and system for rotary machine and storage medium

Country Status (1)

Country Link
CN (1) CN111397896B (en)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112036042B (en) * 2020-09-02 2022-07-26 哈尔滨工程大学 Power equipment abnormality detection method and system based on variational modal decomposition
CN112347588A (en) * 2020-11-26 2021-02-09 中国舰船研究设计中心 Rotary machine fault diagnosis method based on wavelet packet decomposition
CN112541552B (en) * 2020-12-16 2022-04-19 中国计量大学上虞高等研究院有限公司 Air handling unit fault detection and diagnosis method combining DCCNN and LGBM
CN112880726A (en) * 2020-12-31 2021-06-01 北京航天试验技术研究所 Sensor fault diagnosis method and device based on variational modal decomposition sample entropy
CN112945556B (en) * 2021-01-26 2022-07-05 大连海事大学 Bearing fault diagnosis method based on wavelet packet decomposition and optimal rejection classification strategy
CN113076844B (en) * 2021-03-26 2022-05-20 华中科技大学 Method for constructing fault diagnosis model of rotating part and application
CN113739909A (en) * 2021-09-08 2021-12-03 上海海事大学 Ship rotating machinery fault diagnosis method based on full-scale time domain averaging
CN114563181B (en) * 2022-01-10 2023-06-27 浙江工业大学之江学院 Rotary machine fault feature extraction method based on improved variation mode extraction
CN115629930B (en) * 2022-12-23 2023-08-08 北京东远润兴科技有限公司 Fault detection method, device, equipment and storage medium based on DSP system
CN117171625B (en) * 2023-10-23 2024-02-06 云和恩墨(北京)信息技术有限公司 Intelligent classification method and device for working conditions, electronic equipment and storage medium

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107451557A (en) * 2017-07-29 2017-12-08 吉林化工学院 Transmission line short-circuit fault diagnostic method based on experience wavelet transformation and local energy
CN107506709A (en) * 2017-08-15 2017-12-22 南昌航空大学 Aircraft engine rotor system fault-signal characteristics information extraction method
CN109855875A (en) * 2019-01-15 2019-06-07 沈阳化工大学 A kind of rolling bearing operational reliability prediction technique

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101187593A (en) * 2007-12-18 2008-05-28 华南理工大学 Power station steam turbine rotor vibration fault Kolmogorov entropy diagnosis method
FR2999722B1 (en) * 2012-12-19 2022-01-07 Electricite De France LOCATION OF ONE OR MORE FAULTS IN AN ELECTROCHEMICAL ASSEMBLY.
CN103308292B (en) * 2013-05-16 2015-07-29 国家电网公司 Based on the vacuum breaker mechanical state detection method of analysis of vibration signal
CN103345641A (en) * 2013-07-16 2013-10-09 杭州电子科技大学 Hand electromyographic signal motion recognition method based on wavelet entropy and support vector machine
CN103439110B (en) * 2013-07-31 2015-12-09 哈尔滨工程大学 Rolling bearing early-stage weak fault diagnostic method
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
CN110108431B (en) * 2019-05-22 2021-07-16 西安因联信息科技有限公司 Mechanical equipment fault diagnosis method based on machine learning classification algorithm

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107451557A (en) * 2017-07-29 2017-12-08 吉林化工学院 Transmission line short-circuit fault diagnostic method based on experience wavelet transformation and local energy
CN107506709A (en) * 2017-08-15 2017-12-22 南昌航空大学 Aircraft engine rotor system fault-signal characteristics information extraction method
CN109855875A (en) * 2019-01-15 2019-06-07 沈阳化工大学 A kind of rolling bearing operational reliability prediction technique

Also Published As

Publication number Publication date
CN111397896A (en) 2020-07-10

Similar Documents

Publication Publication Date Title
CN111397896B (en) Fault diagnosis method and system for rotary machine and storage medium
Yang et al. Interpretability of deep convolutional neural networks on rolling bearing fault diagnosis
Yiakopoulos et al. Rolling element bearing fault detection in industrial environments based on a K-means clustering approach
CN112347588A (en) Rotary machine fault diagnosis method based on wavelet packet decomposition
CN112633098B (en) Rotary machine fault diagnosis method, system and storage medium
CN111089720B (en) Regularization sparse filtering method suitable for gear fault diagnosis under variable rotating speed
Yu et al. Planetary gear fault diagnosis using stacked denoising autoencoder and gated recurrent unit neural network under noisy environment and time-varying rotational speed conditions
CN111351665B (en) Rolling bearing fault diagnosis method based on EMD and residual error neural network
CN112257530A (en) Rolling bearing fault diagnosis method based on blind signal separation and support vector machine
Zhao et al. Data augmentation via randomized wavelet expansion and its application in few-shot fault diagnosis of aviation hydraulic pumps
Tang et al. A robust deep learning network for low-speed machinery fault diagnosis based on multikernel and RPCA
CN109918417B (en) Time sequence data self-adaptive segmentation, dimension reduction and characterization method based on wavelet transformation and application
Zhou et al. Bearing fault diagnosis based on Gramian angular field and DenseNet
CN112668459A (en) Rolling bearing fault diagnosis method based on supervised contrast learning and convolutional neural network
Zhang et al. Bearing fault diagnosis under various operation conditions using synchrosqueezing transform and improved two-dimensional convolutional neural network
Chen et al. Fault identification method for planetary gear based on DT-CWT threshold denoising and LE
Wang et al. Construction of the efficient attention prototypical net based on the time–frequency characterization of vibration signals under noisy small sample
GS et al. Evaluation of effectiveness of wavelet based denoising schemes using ANN and SVM for bearing condition classification
Yu et al. Machine fault diagnosis based on Gaussian mixture model and its application
Lu et al. Bearing fault diagnosis based on clustering and sparse representation in frequency domain
CN115290326A (en) Rolling bearing fault intelligent diagnosis method
Hu et al. Fault diagnosis based on multi-scale redefined dimensionless indicators and density peak clustering with geodesic distances
Yan et al. Attention mechanism-guided residual convolution variational autoencoder for bearing fault diagnosis under noisy environments
Wang et al. First-order differential filtering spectrum division method and information fusion multi-scale network for fault diagnosis of bearings under different loads
Hu et al. Prior knowledge-based residuals shrinkage prototype networks for cross-domain fault diagnosis

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant