CN116467649A - EEMD permutation entropy and GA-SVM-based turbine expander fault diagnosis method and device - Google Patents
EEMD permutation entropy and GA-SVM-based turbine expander fault diagnosis method and device Download PDFInfo
- Publication number
- CN116467649A CN116467649A CN202310440795.7A CN202310440795A CN116467649A CN 116467649 A CN116467649 A CN 116467649A CN 202310440795 A CN202310440795 A CN 202310440795A CN 116467649 A CN116467649 A CN 116467649A
- Authority
- CN
- China
- Prior art keywords
- svm
- eemd
- imf
- signal
- permutation entropy
- 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
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 45
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 27
- 238000004458 analytical method Methods 0.000 claims abstract description 14
- 230000009467 reduction Effects 0.000 claims abstract description 13
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 230000002159 abnormal effect Effects 0.000 claims abstract description 6
- 238000012549 training Methods 0.000 claims description 30
- 230000006870 function Effects 0.000 claims description 24
- 230000002068 genetic effect Effects 0.000 claims description 17
- 238000005457 optimization Methods 0.000 claims description 16
- 239000013598 vector Substances 0.000 claims description 16
- 238000012795 verification Methods 0.000 claims description 16
- 238000000354 decomposition reaction Methods 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 12
- 238000010606 normalization Methods 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000012706 support-vector machine Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 4
- 238000012360 testing method Methods 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims description 2
- 230000001174 ascending effect Effects 0.000 claims description 2
- 238000010276 construction Methods 0.000 claims description 2
- 238000009826 distribution Methods 0.000 claims description 2
- 230000009977 dual effect Effects 0.000 claims description 2
- 230000007246 mechanism Effects 0.000 claims description 2
- 230000035772 mutation Effects 0.000 claims description 2
- 238000000926 separation method Methods 0.000 claims description 2
- 238000010801 machine learning Methods 0.000 claims 1
- 230000009286 beneficial effect Effects 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000036039 immunity Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 230000035945 sensitivity Effects 0.000 description 2
- 230000036541 health Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000704 physical effect Effects 0.000 description 1
- 238000005381 potential energy Methods 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 238000009423 ventilation Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H17/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves, not provided for in the preceding groups
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Genetics & Genomics (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a turbine expander fault diagnosis method based on EEMD permutation entropy and GA-SVM, which comprises the following steps: s1, acquiring a vibration signal of a turbine expander through signal acquisition equipment, preprocessing the vibration signal, reducing noise of an original vibration signal and improving data reliability; s2, decomposing the preprocessed vibration signal of the turbine expander by using EEMD to obtain a plurality of IMF components and a residual component; s3, calculating a correlation coefficient, determining IMF components with strong correlation with the original data through a threshold value, and forming effective IMF components into corresponding frequency multiplication components. The beneficial effects are that; according to the invention, the original vibration signal is subjected to missing value analysis, abnormal value analysis and noise reduction treatment, so that the disturbance influence of noise can be well avoided, and the reliability of data is improved; according to the invention, the EEMD algorithm is used for decomposing the vibration original signal of the turbine expander, so that the effective information of the vibration signal can be well obtained.
Description
Technical Field
The invention relates to the field of diagnosis devices for related faults of a turboexpander, in particular to a method for realizing rapid diagnosis of the faults of the turboexpander by modeling and deep learning.
Background
In modern industrial systems, the operating state of the critical equipment in the system determines the operating efficiency of the overall industrial system. Therefore, the faults of the equipment need to be recognized and processed in time so as to avoid the stoppage and the loss caused in the production process. The traditional fault diagnosis method is mainly based on an analytical model, the basic idea of the method is to fit the health states of equipment and parts by using a fixed physical formula based on physical properties, the method requires a skilled person to have deep engineering physical knowledge, and the diagnosis quality is lower because the data rule of the equipment is easily ignored by means of excessive physical models. The fault diagnosis method based on data driving and deep learning has become a current hot spot direction, and the running state of the equipment is analyzed based on the running data of the equipment, so that the purpose of fault diagnosis is achieved.
The turbine expander is one of the important parts of energy recovering equipment, low temperature air separating unit and gas separating and liquefying unit, and its operation principle is to convert the potential energy of working medium into output mechanical energy via adiabatic expansion of working medium. The main components of the device comprise a ventilation part, a brake and a machine body. The working medium obtains kinetic energy in the expansion flow part of the turbine expander and has a working wheel shaft to output external work, thereby realizing the aims of recycling waste energy and reducing the internal energy and temperature of the outlet working medium. The turbine expander is used as an important device of a low-temperature system and an energy recovery system, and the working condition of the turbine expander determines the refrigerating capacity of the system and the output of the recovery utilization rate. Failure of the turboexpander can lead to downtime or even damage to the equipment, resulting in reduced efficiency and economic loss of the overall system. The fault diagnosis of the turbine expansion machine is realized by combining a deep learning method, the occurrence of faults is timely alarmed and avoided, and production accidents are reduced.
Disclosure of Invention
Aiming at the problems in the related art, the invention provides a turbine expander fault diagnosis method and device based on EEMD permutation entropy and GA-SVM, which are characterized in that EEMD is utilized to decompose vibration signals of a turbine expander to obtain a plurality of IMF components, and the correlation coefficient between the IMF components and original data is calculated to remove invalid components. Calculating the permutation entropy of the effective IMFs, constructing a multidimensional feature vector of fault features according to the permutation entropy, dividing a data set into a training set and a verification set according to a proportion, and carrying out parameter optimization on the SVM model by utilizing the training set and a genetic algorithm to obtain an optimized model. And finally, inputting the verification set into the optimized SVM model, and identifying and diagnosing faults of the turboexpander.
For this purpose, the invention provides the following technical scheme:
according to a first aspect of an embodiment of the present invention, a turboexpander fault diagnosis method based on EEMD permutation entropy and GA-SVM includes the steps of:
s1, acquiring a vibration signal of a turbine expander through signal acquisition equipment, preprocessing the vibration signal, reducing noise of an original vibration signal and improving data reliability;
s2, decomposing the preprocessed vibration signal of the turbine expander by using EEMD to obtain a plurality of IMF components and a residual component;
s3, calculating a correlation coefficient, determining IMF components with strong correlation with original data through a threshold value, and forming effective IMF components into corresponding frequency multiplication components;
s4, carrying out normalization calculation on the arrangement entropy of different frequency multiplication IMF components, constructing a multidimensional feature vector according to the arrangement entropy, and dividing a data set into a training set and a verification set according to a proportion;
s5, carrying out parameter optimization and training on the SVM model based on a training set and a genetic algorithm to obtain an GS-SVM optimal model, wherein a verification set is used as input of the trained GA-SVM, and a fault diagnosis result of the turboexpander is output;
according to a second aspect of an embodiment of the present invention, a turboexpander fault diagnosis apparatus based on EEMD permutation entropy and GA-SVM includes the following modules:
the acquisition module is used for acquiring vibration data of the turbine expander by utilizing the vibration sensor;
the preprocessing module is used for carrying out missing value analysis, abnormal value analysis and noise reduction processing on the obtained vibration data to obtain a reliable signal;
the calculation module is used for EEMD decomposition of the noise reduction signal, calculating a correlation coefficient and permutation entropy, and selecting IMFs to obtain a multidimensional feature vector;
the training module is used for inputting the training set into the SVM model, training the SVM model based on a genetic algorithm to obtain optimal parameters and obtain an optimal fault diagnosis model;
and the execution module is used for carrying out fault diagnosis on the turbo expander of the GA-SAM model by using the verification set to obtain the fault diagnosis performance of the model.
The invention has the advantages that:
1. according to the invention, the original vibration signal is subjected to missing value analysis, abnormal value analysis and noise reduction treatment, so that the disturbance influence of noise can be well avoided, and the reliability of data is improved;
2. according to the invention, an EEMD algorithm is used for decomposing the vibration original signal of the turbine expander, so that the effective information of the vibration signal can be well obtained;
3. the invention constructs multidimensional feature vectors through permutation entropy, can well improve the sensitivity to signal change and improve the accuracy of fault diagnosis;
4. the multi-dimensional feature vector is constructed through the permutation entropy, the permutation entropy algorithm is simple, the noise immunity is high, and the diagnosis effect is improved;
5. according to the invention, a vibration signal-based turbine expander fault diagnosis model is established by utilizing the SVM, and further, a genetic algorithm is utilized to optimize the penalty factors of the SVM model and the parameters of the kernel function, so that a GA-SVM model with higher accuracy is obtained, and the turbine expander fault is identified and diagnosed.
Drawings
FIG. 1 is a schematic diagram of a turboexpander fault diagnosis method based on EEMD permutation entropy and GA-SVM;
FIG. 2 is a flow chart of a method for diagnosing faults of a turboexpander based on EEMD permutation entropy and GA-SVM;
FIG. 3 is a flowchart of a genetic algorithm in a turboexpander fault diagnosis method based on EEMD permutation entropy and GA-SVM in the present invention;
FIG. 4 is a block diagram of a turboexpander fault diagnosis device based on EEMD permutation entropy and GA-SVM in accordance with the present invention;
FIG. 5 is a sensor mounting diagram of a turboexpander fault diagnosis device based on EEMD permutation entropy and GA-SVM of the present invention;
FIG. 6 is a block diagram of a turboexpander fault diagnosis device based on EEMD permutation entropy and GA-SVM according to the present invention.
Detailed Description
The invention will be described in further detail with reference to the accompanying drawings and specific examples, in order to make the objects, technical solutions and advantages of the invention more apparent. Referring to FIG. 2, a flow chart of a method for diagnosing faults of a turboexpander according to the present invention is shown. The present example is described in terms of a turbine expander bearing failure signal.
S1, collecting vibration signals of a turbine expander bearing through a vibration sensor, wherein the vibration signals are expressed as
And S2, carrying out missing value analysis, outlier processing and noise reduction processing on the vibration signal of the turbine expander bearing to obtain a noise reduction signal x (t).
And S3, decomposing the noise-reduced vibration signal by using the EEMD to obtain a series of IMF components.
Step S31: decomposing the vibration signal x (t) after noise reduction into a plurality of IMF components C by using an EEMD algorithm j The EEMD decomposition step is as follows:
s311 predefines the ensemble-average number N of EEMD algorithms.
S312 predefines the amplitude a of the added normally distributed white noise.
S313, adding normal distribution white noise with the amplitude of A to the original signal to form a new signal. X is x j (t)＝x(t)+n j (t)
x j (t) is the new signal formed, x (t) is the original signal, n j (t) is a normally distributed white noise signal of added amplitude A
S314 will be a new signal x j (t) performing EMD decomposition to obtain each IMF component
Wherein M is the number of IMFs obtained after decomposition, IMFs j，m (t) is the mth IMF obtained in the jth test, the IMF is the eigenmode function, r j，m (t) a residual function representing the average trend of the signal.
S315 repeats steps S31 and S314 with different white noise each time, reaching the ensemble average number of times, obtaining N sets of IMF functions.
S316, carrying out ensemble averaging on the intrinsic mode function IMF obtained in the step S315, wherein the result is as follows:
C j (t) is the jth IMF component obtained after EEMD decomposition. The expression of the decomposed vibration signal x (t) is as follows:
and S4, calculating a correlation coefficient, determining IMF components with strong correlation with the original data through a threshold value, and forming the effective IMF components into corresponding frequency multiplication components.
Step S41: the correlation between the calculated signals (the component with higher correlation when the correlation coefficient is larger than 0.3) is calculated as follows:
wherein x (t) and C j (t) is the original denoising signal and the decomposed IMF component, respectively, and M is the number of decomposed components. The more closely the correlation between the IMF component obtained after EEMD decomposition and the original signal x (t), the more effectively the IMF component, and the larger the correlation coefficient, so that the effective IMF is selected from high to low according to the calculation result of the correlation coefficient.
And S5, carrying out normalization calculation on the arrangement entropy of different frequency multiplication IMF components, constructing a multidimensional feature vector according to the arrangement entropy, and dividing the data set into a training set and a verification set according to the proportion.
Step S51: the permutation entropy normalization calculation process is as follows:
s511, performing phase space reconstruction on each IMF component obtained by selection to obtain the following matrix:
where m is the embedded dimension, τ is the delay factor, k=n- (m-1) τ, t=1, 2, …, k. The matrix has k reconstruction components, each reconstruction component having m-dimensional embedded elements.
S512, the ith component (x (i), x (i+tau), … and x (i+ (m-1) tau)) in the matrix are arranged in ascending order according to the magnitude of the numerical value, so as to obtain:
C j (i+(f 1 -1)τ)≤C j (i+(f 2 -1)τ)≤…≤C j (i+(f m -1)τ)
wherein f 1 ,f 2 ,…,f m Representing the index value of the subscript of each element in the reconstructed component. If there are two or more equal values in the reconstruction component, it is necessary to rely on f 1 ,f 2 Is ordered by size. Satisfy f 1 ＜f 2 Then, the following steps are obtained:
C j (i+(f 1 -1)τ)≤C j (i+(f 2 -1)τ)
s513, for each reconstruction vector C in the reconstruction space j (i) A reconstructed symbol sequence is obtained:
S(l)＝(f 1 ,f 2 ,…f m )
wherein l=1, 2, …, k satisfies k.ltoreq.m ≡! . Each reconstruction vector is an m-dimensional space, mapped to an m-dimensional symbol sequence, and shares m-! An arrangement mode.
S514, calculating the probability P of each m-dimensional symbol sequence 1 ,P 2 ,…,P K Then, according to the definition of the aroma entropy, the sequence { C } j (i) I=1, 2, where, the Permutation Entropy (PE) of g } is calculated as follows:
s515, the permutation entropy is updated into a new permutation entropy after normalization processing, and the new permutation entropy normalization formula is as follows:
0≤H p ＝H p /ln(m！)≤1
step S52, dividing the data set into a training set and a verification set according to the proportion of 7:3;
and S6, carrying out parameter optimization and training on the SVM model by utilizing a training set and a genetic algorithm to obtain the GS-SVM optimal model.
Step S61: the core idea of the support vector machine is to find an optimal hyperplane, separating training samples. The optimal hyperplane is to properly separate the data and maximize the separation. The SVM solves the classification problem by solving an optimization problem.
The SVM construction optimization problem is as follows:
wherein ε i As an introduced relaxation variable; c is penalty factor. Converting the SVM optimal solution problem into a dual problem by using a Lagrangian multiplier method, and solving the maximum value under the constraint condition:
where α is the Lagrangian multiplier, k (x i ,x j ) Is a kernel function; for complex data conditions, the kernel functions map the data to a high-dimensional space, and different kernel functions have certain influence on the classification capability of the SVM. Wherein, the kernel function adopts a Gaussian radial basis function, and the expression is as follows:
k(x i ,x j )＝exp(-g||x i -x j || 2 )
the optimization function corresponding to the support vector machine is:
wherein alpha is i And b is a bias variable, which is a Lagrangian multiplier.
The performance of the SVM model is directly related to the parameters thereof, and referring to FIG. 3, the invention adopts a genetic algorithm (Genetic Algorithm, GA) to optimize the kernel function parameter g and the penalty factor C in the SVM model.
Step S62: the genetic algorithm is a random global search optimization method, and mainly utilizes a coding method and a propagation mechanism to realize optimization of SVM parameters until the parameters meet termination conditions or reach the maximum iteration times. The main operations include: selection operation, crossover operation and mutation operation.
The GA optimizes the SVM parameters as follows:
s621, initializing parameters;
s622, binary coding is carried out on a nuclear parameter g and a punishment factor C of the SVM model to form an initial population;
s623, calculating a fitness value, judging whether a termination condition is met by the GA, and decoding and outputting optimal parameters g and C if the termination condition is met.
Otherwise, selecting, crossing and mutating the population to form a new population;
s624, analyzing the offspring population, calculating a new fitness value, and judging whether termination conditions are met by the GA; step S624 is repeated until the last optimal parameter is generated.
And S7, taking the verification set as input of the GA-SVM after training is finished, outputting a bearing fault diagnosis result of the turboexpander, and judging the fault state of the bearing according to the classification result. And comparing the result with the truly classified labels through the test of the verification set, and calculating the accuracy of model classification to obtain the truly performance of the model.
Referring to fig. 4, the present embodiment further provides a turboexpander fault diagnosis apparatus based on EEMD permutation entropy and GA-SVM, which corresponds to a virtual apparatus based on EEMD permutation entropy and GA-SVM, which includes:
an acquisition module for acquiring vibration data of the turboexpander by using a vibration sensor, the installation position of which is referred to in fig. 5;
the preprocessing module is used for carrying out missing value analysis, abnormal value analysis and noise reduction processing on the obtained vibration data to obtain a reliable signal;
the calculation module is used for EEMD decomposition of the noise reduction signals, calculating correlation coefficients, selecting IMFs, and calculating permutation entropy to obtain multidimensional feature vectors;
the training module is used for inputting the training set into the SVM model, training the SVM model based on a genetic algorithm to obtain optimal parameters and obtain an optimal fault diagnosis model;
and the execution module is used for using the verification set for fault diagnosis of the turboexpander of the GA-SVM model to obtain the fault diagnosis performance of the model.
Referring to fig. 6, a fault diagnosis device for a turboexpander based on EEMD permutation entropy and GA-SVM model is divided into five modules, an acquisition module acquires vibration signals through a vibration sensor, data is transmitted to a visualization system through an OPC protocol, a preprocessing module can preprocess the data by using a data processing algorithm in a computer, a calculation module, a training module and an execution module complete processing calculation through algorithm codes of the computer, and the performance of the model of the invention is judged through classification graphic and the size of classification index. The functional units in the embodiment of the device of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
In summary, by means of the above technical solution of the present invention, the missing value analysis, the outlier analysis and the noise reduction processing are performed on the original vibration signal, which can well avoid the disturbance effect of noise, and improve the reliability of data; according to the invention, an EEMD algorithm is used for decomposing the vibration original signal of the turbine expander, so that the effective information of the vibration signal can be well obtained; the invention constructs multidimensional feature vectors through permutation entropy, can well improve the sensitivity to signal change and improve the accuracy of fault diagnosis; the multi-dimensional feature vector is constructed through the permutation entropy, the permutation entropy algorithm is simple, the noise immunity is high, and the diagnosis effect is improved; according to the invention, a vibration signal-based turbine expander fault diagnosis model is established by utilizing the SVM, and further, a genetic algorithm is utilized to optimize the penalty factors of the SVM model and the parameters of the kernel function, so that a GA-SVM model with higher accuracy is obtained, and the turbine expander fault is identified and diagnosed.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, alternatives, and improvements that fall within the spirit and scope of the invention.
Claims (7)
1. A turbine expander fault diagnosis method based on EEMD permutation entropy and GA-SVM is characterized in that: the method comprises the following steps:
s1, acquiring a vibration signal of a turbine expander through signal acquisition equipment, preprocessing the vibration signal, reducing noise of an original vibration signal and improving data reliability;
s2, decomposing the preprocessed vibration signal of the turbine expander by using EEMD to obtain a plurality of IMF components and a residual component;
s3, calculating a correlation coefficient, determining IMF components with strong correlation with the original data through a threshold value, and forming effective IMF components into corresponding frequency multiplication components;
s4, carrying out normalization calculation on the arrangement entropy of different frequency multiplication IMF components, constructing a multidimensional feature vector according to the arrangement entropy, and dividing a data set into a training set and a verification set according to a proportion;
and S5, carrying out parameter optimization and training on the SVM model based on a training set and a genetic algorithm, obtaining a GS-SVM optimization model, taking a verification set as input of the trained GA-SVM, and outputting a fault diagnosis result of the turboexpander.
2. The turboexpander fault diagnosis method based on EEMD permutation entropy and GA-SVM according to claim 1, wherein the data acquisition and preprocessing process in S1 comprises: collecting vibration data of the turbine expander through a vibration sensor; and carrying out missing value analysis, abnormal value processing and noise reduction processing on the obtained vibration data.
3. The method for diagnosing a failure of a turboexpander based on EEMD permutation entropy and GA-SVM as set forth in claim 1, wherein the decomposing the preprocessed vibration signal with EEMD in S2 to obtain a plurality of IMF components and a residual component includes:
the EEMD decomposition process is as follows:
s21, predefining the overall average number N of EEMD algorithm and the amplitude A of the added normal distribution white noise;
s22, adding normal distributed white noise with the amplitude of A into the original signal to form a new signal:
x j (t)＝x(t)+n j (t)
wherein x is j (t) is the new signal formed, x (t) is the original signal, n j (t) is a normally distributed white noise signal of added amplitude A;
s23, new signal x j (t) performing EMD decomposition to obtain each IMF component as follows:
wherein M is the number of IMFs obtained after decomposition, IMFs j,m (t) is the mth IMF obtained in the jth test, the IMF is the eigenmode function, r j,m (t) a residual function representing the average trend of the signal;
s24, repeating the steps S23 and S24 with different white noise each time to reach the total average times, and obtaining N groups of IMF functions;
s25, carrying out ensemble averaging on the intrinsic mode function IMF obtained in the step S24, wherein the result is as follows:
C j (t) is the jth IMF component obtained after EEMD decomposition.
4. The turboexpander fault diagnosis method based on EEMD permutation entropy and GA-SVM according to claim 1, wherein calculating the correlation coefficient in S3, selecting effective IMF components comprises:
x (t) is an original signal, IMF is a signal obtained after decomposition, and then the correlation coefficient is calculated as:
wherein x (t) and C j (t) the original de-noised signal and the decomposed IMF, respectivelyThe component M is the number of the components obtained by decomposition; the more closely the correlation between the IMF component obtained after EEMD decomposition and the original signal x (t), the more effectively the IMF component, and the larger the correlation coefficient, so that the effective IMF is selected from high to low according to the calculation result of the correlation coefficient.
5. The turboexpander fault diagnosis method based on EEMD permutation entropy and GA-SVM as set forth in claim 1, wherein said step S4 of normalizing permutation entropy of different frequency multiplication IMF components, constructing multidimensional feature vectors according to permutation entropy, and dividing data set into training set and verification set according to proportion, comprises:
the permutation entropy normalization calculation process is as follows:
s41, carrying out phase space reconstruction on each IMF component obtained through selection to obtain the following matrix:
where m is the embedded dimension, τ is the delay factor, k=n- (m-1) τ, t=1, 2, …, k; the matrix has k reconstruction components, and each reconstruction component has m-dimensional embedded elements;
s42, the ith component (x (i), x (i+tau), … and x (i+ (m-1) tau)) in the matrix are arranged in ascending order according to the magnitude of the numerical value, so as to obtain:
C j (i+(f 1 -1)τ)≤C j (i+(f 2 -1)τ)≤…≤C j (i+(f m -1)τ)
wherein f 1 ,f 2 ,…,f m Index values representing subscripts of the elements in the reconstructed component; if there are two or more equal values in the reconstruction component, it is necessary to rely on f 1 ,f 2 Ordering the sizes of (3); satisfy f 1 ＜f 2 Then, the following steps are obtained: c (C) j (i+(f 1 -1)τ)≤C j (i+(f 2 -1)τ)
S43, for each reconstruction vector C in the reconstruction space j (i) A reconstructed symbol sequence is obtained:
S(l)＝(f 1 ,f 2 ,…f m )
wherein l=1, 2, …, k satisfies k.ltoreq.m ≡! The method comprises the steps of carrying out a first treatment on the surface of the Each reconstruction vector is an m-dimensional space, mapped to an m-dimensional symbol sequence, and shares m-! A seed arrangement mode;
s44, calculating the probability P of each m-dimensional symbol sequence 1 ,P 2 ,…,P K Then, according to the definition of the aroma entropy, the sequence { C } j (i) I=1, 2, where, the Permutation Entropy (PE) of g } is calculated as follows:
s45, after normalization processing is carried out on the permutation entropy, updating the permutation entropy into a new permutation entropy, wherein a new permutation entropy normalization formula is as follows: h is more than or equal to 0 p ＝H p And/ln (m|) is less than or equal to 1, and the data set is divided into a training set and a verification set according to the proportion of 7:3.
6. The method for diagnosing a failure of a turboexpander based on EEMD permutation entropy and GA-SVM as set forth in claim 1, wherein in S5, parameter optimization and training are performed on an SVM model based on a training set and a genetic algorithm to obtain an optimal model of the GS-SVM, the verification set is used as an input of the GA-SVM after training is completed, and a failure diagnosis result of the turboexpander is output, comprising:
the support vector machine (Support Vector Machines, SVM) is a branch of machine learning, the core idea of which is to find an optimal hyperplane, separating training samples; the optimal hyperplane is to correctly separate the data and maximize the separation; the SVM solves the classification problem by solving the optimization problem;
the SVM construction optimization problem is as follows:
wherein ε i As an introduced relaxation variable; c is penalty factor; solving problem of SVM optimal solution by Lagrangian multiplier methodConverting into a dual problem, and solving the maximum value under the constraint condition:
where α is the Lagrangian multiplier, k (x i ,x j ) Is a kernel function; for complex data conditions, the kernel functions map the data to a high-dimensional space, and different kernel functions have different influences on the classification capacity of the SVM; wherein, the kernel function adopts a Gaussian radial basis function, and the expression is as follows:
k(x i ,x j )＝exp(-g||x i -x j || 2 )
the optimization function corresponding to the support vector machine is:
wherein alpha is i B is a Lagrangian multiplier, b is a bias variable;
the performance of the SVM model is directly related to the parameters thereof, and the genetic algorithm (Genetic Algorithm, GA) is adopted to optimize the kernel function parameter g and the penalty factor C in the SVM model;
the genetic algorithm is a random global search optimization method, and mainly utilizes a coding method and a propagation mechanism to realize optimization of SVM parameters until the parameters meet termination conditions or reach the maximum iteration times; the main operations include: a selection operation, a crossover operation and a mutation operation;
the GA optimizes the SVM parameters as follows:
s51, initializing parameters;
s52, binary coding is carried out on the kernel parameter g and the penalty factor C of the SVM model to form an initial population;
s53, calculating a fitness value, judging whether a termination condition is met by the GA, and decoding to output optimal parameters g and C if the termination condition is met; otherwise, selecting, crossing and mutating the population to form a new population;
s54, analyzing the offspring population, calculating a new fitness value, and judging whether termination conditions are met by the GA; the steps S54 are repeated until the last optimal parameters are generated.
7. A turboexpander fault diagnosis device based on EEMD permutation entropy and GA-SVM for implementing the diagnosis method of claim 1, comprising:
the acquisition module is used for acquiring vibration data of the turbine expander by utilizing the vibration sensor;
the preprocessing module is used for carrying out missing value analysis, abnormal value analysis and noise reduction processing on the obtained vibration data to obtain a reliable signal;
the calculation module is used for EEMD decomposition of the noise reduction signals, calculating correlation coefficients, selecting IMFs, and calculating permutation entropy to obtain multidimensional feature vectors;
the training module is used for inputting the training set into the SVM model, training the SVM model based on a genetic algorithm to obtain optimal parameters and obtain an optimal fault diagnosis model;
and the execution module is used for carrying out fault diagnosis on the turbo expander of the GA-SAM model by using the verification set to obtain the fault diagnosis performance of the model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310440795.7A CN116467649A (en) | 2023-04-23 | 2023-04-23 | EEMD permutation entropy and GA-SVM-based turbine expander fault diagnosis method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310440795.7A CN116467649A (en) | 2023-04-23 | 2023-04-23 | EEMD permutation entropy and GA-SVM-based turbine expander fault diagnosis method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116467649A true CN116467649A (en) | 2023-07-21 |
Family
ID=87175007
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310440795.7A Pending CN116467649A (en) | 2023-04-23 | 2023-04-23 | EEMD permutation entropy and GA-SVM-based turbine expander fault diagnosis method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116467649A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117093854A (en) * | 2023-10-19 | 2023-11-21 | 安徽建筑大学 | Transformer mechanical fault diagnosis method, equipment and storage medium |
| CN117633696A (en) * | 2024-01-26 | 2024-03-01 | 深圳市智德森水务科技有限公司 | Hydraulic engineering construction monitoring method and system based on BIM model |
| CN117909795A (en) * | 2024-01-31 | 2024-04-19 | 昆明理工大学 | Water pump unit fault diagnosis method based on EEMD (EEMD) and LCTDBO-BP fusion |
-
2023
- 2023-04-23 CN CN202310440795.7A patent/CN116467649A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117093854A (en) * | 2023-10-19 | 2023-11-21 | 安徽建筑大学 | Transformer mechanical fault diagnosis method, equipment and storage medium |
| CN117093854B (en) * | 2023-10-19 | 2024-02-09 | 安徽建筑大学 | Transformer mechanical fault diagnosis method, equipment and storage medium |
| CN117633696A (en) * | 2024-01-26 | 2024-03-01 | 深圳市智德森水务科技有限公司 | Hydraulic engineering construction monitoring method and system based on BIM model |
| CN117633696B (en) * | 2024-01-26 | 2024-04-26 | 深圳市智德森水务科技有限公司 | Hydraulic engineering construction monitoring method and system based on BIM model |
| CN117909795A (en) * | 2024-01-31 | 2024-04-19 | 昆明理工大学 | Water pump unit fault diagnosis method based on EEMD (EEMD) and LCTDBO-BP fusion |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116467649A (en) | EEMD permutation entropy and GA-SVM-based turbine expander fault diagnosis method and device | |
| CN105738109B (en) | Bearing fault classification diagnosis method based on rarefaction representation and integrated study | |
| Huang et al. | Memory residual regression autoencoder for bearing fault detection | |
| CN105678343B (en) | Hydropower Unit noise abnormality diagnostic method based on adaptive weighted group of sparse expression | |
| CN113505655B (en) | Intelligent bearing fault diagnosis method for digital twin system | |
| CN117034123B (en) | Fault monitoring system and method for fitness equipment | |
| CN110083593B (en) | Power station operation parameter cleaning and repairing method and repairing system | |
| CN115600140A (en) | Fan variable pitch system fault identification method and system based on multi-source data fusion | |
| CN115587335A (en) | Training method of abnormal value detection model, abnormal value detection method and system | |
| CN110766215B (en) | Wind power climbing event prediction method based on feature adaptive selection and WDNN | |
| CN116374645A (en) | Running state monitoring system and method for bucket wheel machine | |
| CN114595635A (en) | Feature selection method, system and equipment for main steam temperature data of thermal power generating unit | |
| Cabrera et al. | Adversarial fault detector guided by one-class learning for a multistage centrifugal pump | |
| Hu et al. | Mutual information-based feature disentangled network for anomaly detection under variable working conditions | |
| Hu et al. | Root cause identification of industrial alarm floods using word embedding and few-shot learning | |
| CN117421655A (en) | Industrial Internet data stream anomaly detection method and system | |
| Yuan et al. | Bearing Fault Diagnosis Based on Improved Locality‐Constrained Linear Coding and Adaptive PSO‐Optimized SVM | |
| CN116910657A (en) | Fault diagnosis method and equipment based on unsupervised learning | |
| Pei et al. | Degradation state identification for hydraulic pumps based on multi-scale ternary dynamic analysis, NSGA-II and SVM | |
| Xu et al. | Fault diagnosis on imbalanced data using an adaptive cost-sensitive multiscale attention network | |
| CN115081555A (en) | Anomaly detection method and device based on generation countermeasure and bidirectional cyclic neural network | |
| CN113505817A (en) | Self-adaptive weighting training method for bearing fault diagnosis model samples under unbalanced data | |
| Lv et al. | A spatial–temporal variational graph attention autoencoder using interactive information for fault detection in complex industrial processes | |
| CN112836570A (en) | Equipment anomaly detection method utilizing Gaussian noise | |
| CN111967593A (en) | Method and system for processing abnormal data based on modeling |
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 |