CN113627358A - Multi-feature fusion turnout intelligent fault diagnosis method, system and equipment - Google Patents

Multi-feature fusion turnout intelligent fault diagnosis method, system and equipment Download PDF

Info

Publication number
CN113627358A
CN113627358A CN202110930537.8A CN202110930537A CN113627358A CN 113627358 A CN113627358 A CN 113627358A CN 202110930537 A CN202110930537 A CN 202110930537A CN 113627358 A CN113627358 A CN 113627358A
Authority
CN
China
Prior art keywords
turnout
fault diagnosis
oil pressure
svm
feature
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110930537.8A
Other languages
Chinese (zh)
Inventor
赵永军
唐军挪
唐天翼
卢瑞冰
张峰
闫卫刚
张嵩
赵勇
许剑财
冯起斌
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.)
Taiyuan Jingfeng Railway Equipment Manufacturing Co ltd
Original Assignee
Taiyuan Jingfeng Railway Equipment Manufacturing Co ltd
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 Taiyuan Jingfeng Railway Equipment Manufacturing Co ltd filed Critical Taiyuan Jingfeng Railway Equipment Manufacturing Co ltd
Priority to CN202110930537.8A priority Critical patent/CN113627358A/en
Publication of CN113627358A publication Critical patent/CN113627358A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • 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
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • 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
    • 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)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Biophysics (AREA)
  • Health & Medical Sciences (AREA)
  • Signal Processing (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)
  • Train Traffic Observation, Control, And Security (AREA)

Abstract

The invention discloses a switch intelligent fault diagnosis method, a system and equipment with multi-feature fusion, wherein the method comprises the following steps: collecting oil pressure signal data of an electro-hydraulic switch machine; decomposing the collected oil pressure signal data by using a MEEMD method to obtain a series of intrinsic mode functions IMF with different time scales, introducing a correlation coefficient method, screening IMF components which truly reflect the original oil pressure signals, calculating sample entropy and energy entropy of the IMF components, performing feature fusion on the obtained information entropy through a kernel principal component analysis method, and constructing a turnout fault diagnosis feature data set by using the obtained result; and (3) optimizing a Support Vector Machine (SVM) by adopting a longicorn stigma search algorithm and a differential evolution algorithm in combination with the constructed data set sample, establishing an improved SVM turnout fault diagnosis model, and identifying and classifying different operation states of the turnout to obtain a turnout fault diagnosis result. The invention has high precision, high accuracy, high efficiency and low cost for classifying the faults of the high-speed railway switch machine equipment.

Description

Multi-feature fusion turnout intelligent fault diagnosis method, system and equipment
Technical Field
The invention relates to the technical field of turnout fault diagnosis of high-speed railways, in particular to a turnout intelligent fault diagnosis method, a turnout intelligent fault diagnosis system and turnout intelligent fault diagnosis equipment with multi-feature fusion.
Background
With the increase of the running speed of high-speed trains in China, ZY series turnout conversion equipment is widely applied to various large railway trunks. The switch conversion equipment is an important railway signal basic equipment, and the running state of the switch conversion equipment is closely related to the safe running and the transportation efficiency of the train. The turnout switching equipment requires the service life of the upper railway to be 15 years, and once the operation state of the turnout switching equipment is deteriorated or fault hidden trouble is not timely and effectively eliminated during maintenance, the normal driving order is inevitably influenced, and even great life and property loss is possibly caused. Therefore, in order to improve the working safety and reliability of the turnout intelligent fault diagnosis system, the intelligent turnout fault diagnosis method based on the data analysis processing technology has great significance to the long-term operation and development of railways.
At present, the fault identification work of railway signal turnouts and turnout conversion equipment in China is mainly completed based on a monitoring system threshold judgment and manual browsing assistance mode, the mode is small in coverage range and low in efficiency, and is affected by uneven manual experience, and the fault judgment result is difficult to standardize. In recent years, with the rapid development of artificial intelligence technology, intelligent methods have replaced human beings to complete some high-strength and high-repeatability operations, and there are many fault analysis methods for turnout, such as: expert systems, knowledge engineering, fuzzy logic, Support Vector Machines (SVMs), neural networks, and the like, each having advantages and disadvantages. At present, the fault diagnosis method for railway turnouts still has the following defects:
(1) researchers mainly combine the current and the power curve in the turnout action process to realize fault diagnosis, and the method has the defects that the power curve is obtained by the current at different moments, and factors influencing the current are more, such as resistance change of a motor, operation conditions, voltage fluctuation and the like, so that the error caused by fault judgment through the power curve is larger.
(2) At present, a fault diagnosis standard of a system is not established in ZY series turnout conversion equipment, the fault diagnosis of a switch machine mainly depends on some simple calculations on original data to obtain a fault diagnosis characteristic set which is used as a fault diagnosis characteristic parameter, the change of a signal in a certain time domain or frequency domain and the change information of the signal in a local mutation position are not considered in the fault diagnosis characteristic parameters, and meanwhile, the problem of single fault characteristic extraction exists, and the fault diagnosis of the turnout is not facilitated.
(3) Aiming at the point switch machine fault diagnosis research method, a perfect core algorithm is not formed, and the existing algorithm has the problems of low fault diagnosis accuracy, long training time and the like.
(4) Due to the particularity of the field device, the fault is difficult to reproduce by 100%, so that the method for diagnosing the fault only depends on hardware equipment, and the accuracy is low.
Therefore, the existing turnout fault diagnosis method still has the inconvenience and defects, and needs to be further improved. How to create a novel turnout intelligent fault diagnosis method, system and equipment with multi-feature fusion, signal features are extracted from oil pressure signal data of a turnout conversion system, and then a SVM model optimized by intelligent optimization algorithm parameters is adopted for fault mode recognition.
Disclosure of Invention
The invention provides a turnout intelligent fault diagnosis method with multi-feature fusion, which is used for extracting signal features from oil pressure signal data of a turnout conversion system and further adopting an SVM model optimized by intelligent optimization algorithm parameters to identify a fault mode.
In order to solve the technical problem, the invention provides a switch intelligent fault diagnosis method with multi-feature fusion, which comprises the following steps:
(1) collecting oil pressure signal data of a turnout electro-hydraulic switch machine; (2) decomposing the collected oil pressure signal data by using a MEEMD method to obtain a series of intrinsic mode functions IMF with different time scales, introducing a correlation coefficient method, screening IMF components which truly reflect the original oil pressure signals, calculating sample entropy and energy entropy, performing feature fusion on the obtained information entropy by using a kernel principal component analysis method, and constructing a turnout fault diagnosis feature data set by using the obtained result; (3) optimizing a Support Vector Machine (SVM) by adopting a Tianniu beard search algorithm BAS and a differential evolution algorithm DE in combination with the constructed turnout fault diagnosis feature data set sample, and establishing an improved SVM turnout fault diagnosis model; (4) and identifying and classifying different running states of the turnout by adopting the constructed improved SVM turnout fault diagnosis model to obtain a turnout fault diagnosis result.
In a further improvement, the acquisition of the oil pressure signal data in the step (1) is realized by oil pressure detection sensors arranged on left and right oil cavities of a starting oil cylinder in the switch machine.
In a further improvement, the decomposition process of the MEEMD method on the oil pressure signal data in the step (2) is as follows:
Figure BDA0003211148870000031
wherein x (t) represents original oil pressure data, [ c ]k(t)]Represents the jth IMF component of the ith signal, and r (t) represents the residual signal.
The decomposition step of the MEEMD method on the oil pressure signal data in the step (2) comprises the following steps:
A1. adding two sets of white noise n with equal amplitude and standard deviation and opposite direction into original signal x (t)i(t)；
Figure BDA0003211148870000032
In the formula, aiRepresents the standard deviation of the ith white noise;
A2. decomposition by EEMD
Figure BDA0003211148870000041
And
Figure BDA0003211148870000042
obtaining Ne IMF components per noise signal;
Figure BDA0003211148870000043
Figure BDA0003211148870000044
in the formula (I), the compound is shown in the specification,
Figure BDA0003211148870000045
representing the jth IMF component of the ith signal.
A3. The average value of the jth IMF component of the ith signal is solved and is recorded as cj(t) subjecting
Figure BDA0003211148870000046
And
Figure BDA0003211148870000047
taking the average, i.e.
Figure BDA0003211148870000048
A4. For calculationObtained cj(t) EMD decomposition to obtain:
Figure BDA0003211148870000049
further improving, in the step (2), by introducing a correlation coefficient method, an IMF component with a correlation coefficient larger than 0.1 with an original signal is screened out, and the sample entropy and the energy entropy of the IMF component are calculated;
wherein, the calculation formula of the sample entropy is as follows:
Figure BDA00032111488700000410
in the formula, Bm(r)、Bm+1(r) represents the number of the maximum absolute difference of the samples of the m-dimensional and m + 1-dimensional data smaller than or equal to the similarity tolerance r, respectively;
the calculation formula of the energy entropy is as follows:
Figure BDA00032111488700000411
in the formula, PiMeaning the energy of the i-th order IMF component contributes to the total energy.
Further improved, the specific steps of performing feature fusion on the obtained information entropy by the kernel principal component analysis method in the step (2) are as follows:
B1. selecting a kernel function K;
B2. calculating a kernel function matrix:
Ki,j＝K(xi,xj)，i,j＝1,2,…,n
B3. carrying out normalization processing on the kernel function matrix:
K’←K-InK-KIn+InKIn
B4. calculating the eigenvalue λ of the kernel function matrix KiAnd feature vector α:
[λ,γ]＝eig(K)
λ＝(λ1,λ2,…,λn)
α＝[α1,α2,…,αn]
B5. calculating the eigenvalue lambdaiCorresponding contribution rate, then normalization processing:
Figure BDA0003211148870000051
B6. the original signal xiProjecting the k principal components to a feature vector alpha to extract k principal components:
Figure BDA0003211148870000052
Figure BDA0003211148870000053
in the formula (I), the compound is shown in the specification,
Figure BDA0003211148870000054
which represents the k-th principal component,
Figure BDA0003211148870000055
representing the projected matrix.
Further improvement, the method for establishing the improved SVM turnout fault diagnosis model in the step (3) specifically comprises the following steps: dividing the constructed turnout fault diagnosis characteristic data set sample into a training set and a testing set according to the ratio of 7:3, training an original SVM model by using training set data, and optimizing two parameter values of a penalty factor c and a kernel function parameter g in the SVM model by using a longicorn algorithm BAS and a differential evolution algorithm DE respectively on the basis of the original SVM model so as to obtain an optimized improved SVM turnout fault diagnosis model with high fault diagnosis accuracy.
Further improvement, the different operation states of the turnout in the step (4) comprise four states of normal operation of a point switch, no unlocking of the turnout jamming, no locking of the turnout jamming and abnormal resistance in the conversion process.
As a further improvement, the invention also provides a switch intelligent fault diagnosis system with multi-feature fusion, which comprises: the oil pressure monitoring module is used for acquiring oil pressure signal data of the turnout electro-hydraulic switch machine; the turnout fault diagnosis characteristic data set construction module is used for decomposing and processing the collected oil pressure signal data by using an MEEMD method to obtain a series of intrinsic mode functions IMF with different time scales, introducing a correlation coefficient method, screening IMF components which truly reflect original oil pressure signals, calculating sample entropy and energy entropy of the IMF components, then performing characteristic fusion on the obtained information entropy by using a kernel principal component analysis method, and constructing a turnout fault diagnosis characteristic data set by using the obtained result; the improved SVM turnout fault diagnosis model building module is used for optimizing a Support Vector Machine (SVM) by combining a built turnout fault diagnosis feature data set sample and adopting a Tianniu beard search algorithm BAS and a differential evolution algorithm DE to build an improved SVM turnout fault diagnosis model; and the turnout fault diagnosis result output module is used for identifying and classifying different operation states of the turnout by adopting the constructed improved SVM turnout fault diagnosis model and outputting a turnout fault diagnosis result.
As a further improvement, the present invention further provides a computer device, which includes a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned switch intelligent fault diagnosis method with multi-feature fusion.
After adopting such design, the invention has at least the following advantages:
the invention relates to a turnout intelligent fault diagnosis method with multi-feature fusion.A MEEMD method is used for decomposing turnout conversion oil pressure data to obtain a series of intrinsic mode functions IMF with different time scales; then, a correlation coefficient method is introduced, IMF components capable of truly reflecting the original oil pressure signals are screened out, and sample entropy and energy entropy of the IMF components are calculated; then, performing feature fusion on the obtained information entropy through a Kernel Principal Component Analysis (KPCA) method, and constructing a turnout fault diagnosis feature set according to the obtained result; and finally, optimizing SVM parameters by adopting a Tianniu whisker search algorithm BAS and a differential evolution algorithm DE in combination with the turnout fault diagnosis characteristic set sample, constructing a turnout fault diagnosis model of the improved SVM, identifying and classifying four states of the turnout, and accurately outputting a turnout fault result. The method has high precision in classifying the faults of the high-speed railway switch machine equipment, and has the advantages of high efficiency and low cost.
Drawings
The foregoing is only an overview of the technical solutions of the present invention, and in order to make the technical solutions of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and the detailed description.
Fig. 1 is a pressure diagram of an electro-hydraulic switch machine in a normal operating state.
Fig. 2 is a pressure curve diagram of the turnout jam in the unlocked state.
Fig. 3 is a pressure curve diagram of the turnout in a locked state.
Fig. 4 is a pressure graph showing an abnormal resistance state during the transition.
FIG. 5 is a waveform diagram of the MEEMD decomposition result of the normal state signal of the electro-hydraulic switch machine.
Fig. 6 is a schematic block diagram of the design of the fault diagnosis method of the present invention.
Fig. 7 is a logic flow diagram of the BAS-SVM algorithm in the fault diagnosis method of the present invention.
Fig. 8 is a logic flow diagram of the DE-SVM algorithm in the fault diagnosis method of the present invention.
Detailed Description
By analyzing the existing turnout fault diagnosis method, the invention selects collected oil pressure data to diagnose the fault of the turnout switch machine in consideration of the fact that the system oil pressure is relatively stable in the working process of the existing electro-hydraulic turnout switch equipment, and factors influencing the oil pressure are relatively few and mainly related to internal power, additional force and conversion resistance. Specific examples are as follows.
This many characteristics of this embodiment fuse's switch intelligence fault diagnosis system includes:
the oil pressure monitoring module is used for acquiring oil pressure signal data of the turnout electro-hydraulic switch machine;
the turnout fault diagnosis characteristic data set construction module is used for decomposing and processing the collected oil pressure signal data by using an MEEMD method to obtain a series of intrinsic mode functions IMF with different time scales, introducing a correlation coefficient method, screening IMF components which truly reflect original oil pressure signals, calculating sample entropy and energy entropy of the IMF components, then performing characteristic fusion on the obtained information entropy by using a kernel principal component analysis method, and constructing a turnout fault diagnosis characteristic data set by using the obtained result;
the improved SVM turnout fault diagnosis model building module is used for optimizing a Support Vector Machine (SVM) by combining a built turnout fault diagnosis feature data set sample and adopting a Tianniu beard search algorithm BAS and a differential evolution algorithm DE to build an improved SVM turnout fault diagnosis model;
and the turnout fault diagnosis result output module is used for identifying and classifying different operation states of the turnout by adopting the constructed improved SVM turnout fault diagnosis model and outputting a turnout fault diagnosis result.
The switch intelligent fault diagnosis method based on multi-feature fusion implemented by the multi-feature fusion switch intelligent fault diagnosis system in the embodiment comprises the following steps:
the method comprises the following steps: collecting oil pressure signal data of the railway electro-hydraulic switch machine.
In the data acquisition process, the oil pressure signal data acquisition is realized through the oil pressure monitoring module arranged in the switch machine equipment. The interior of the ZY series electro-hydraulic switch machine mainly comprises a three-phase alternating current motor, a coupling, an oil pump and a hydraulic cylinder. In the embodiment, the oil pressure measuring points are arranged on the left oil cavity and the right oil cavity of the starting oil cylinder which can better reflect the working state of the actual oil circuit system, and the oil pressure monitoring sensors are respectively installed, so that the working pressure of the oil circuit system can be monitored and collected.
According to the conversion process state of the electro-hydraulic switch machine, oil pressure data of the switch machine in four states of normal state, no unlocking of turnout blockage, no locking of turnout blockage and abnormal resistance in the conversion process are collected, and in order to simplify different state names, a fault is used uniformly0、fault1、fault2、fault3Instead of this.
In the data collection process, firstly, pressure signals under the normal state of the switch machine are collected, as shown in fig. 1, the measurement precision is 0.1MPa, the monitoring system collects oil pressure data at the collection frequency of 50Hz, the simulation of the conversion process is carried out according to the four working states, and meanwhile, complete oil pressure signals under the four states are collected. As shown in figure 2, the pressure curve diagram of the turnout in the unlocked state is shown; FIG. 3 shows a pressure curve diagram of a turnout in a locked state; FIG. 4 is a pressure profile for a transition in the presence of abnormal resistance.
Step two: and extracting the data characteristics of the switch machine oil pressure signal.
In the embodiment, an integrated empirical mode decomposition (MEEMD) method is adopted to decompose oil pressure data in a conversion process to obtain a series of Intrinsic Mode Functions (IMFs) with different time scales, then a correlation coefficient method is introduced to screen out IMF components capable of truly reflecting original oil pressure signals, sample entropy and energy entropy of the IMF components are calculated, then feature fusion is carried out on obtained information entropy through a Kernel Principal Component Analysis (KPCA), and a turnout fault diagnosis feature set is constructed on obtained results.
Specifically, first, the collected oil pressure data of the switch machine in the conversion process is decomposed by the MEEMD algorithm to obtain 8 IMF components, and a graph of the IMF components obtained by the oil pressure signals through the MEEMD decomposition is shown in fig. 5, taking the signals of the normal operation condition as an example.
The MEEMD algorithm comprises the following decomposition steps:
(1) adding two sets of white noise n with equal amplitude and standard deviation and opposite direction into original signal x (t)i(t)。
Figure BDA0003211148870000101
In the formula, aiRepresents the standard deviation of the ith white noise;
(2) decomposition by EEMD
Figure BDA0003211148870000102
And
Figure BDA0003211148870000103
each noise signal results in Ne IMF components.
Figure BDA0003211148870000104
Figure BDA0003211148870000105
In the formula (I), the compound is shown in the specification,
Figure BDA0003211148870000106
representing the jth IMF component of the ith signal.
(3) The average value of the jth IMF component of the ith signal is solved and is recorded as cj(t) of (d). Will be provided with
Figure BDA0003211148870000107
And
Figure BDA0003211148870000108
taking the average, i.e.
Figure BDA0003211148870000109
(4) C calculated according to the formula (2-4)j(t) may not be the standard IMF component and there may be modal splitting phenomena, so c is neededj(t) EMD decomposition, taking the first order component as an example, namely:
Figure BDA00032111488700001010
Figure BDA00032111488700001011
that is, the decomposition process of the MEEMD algorithm can be simplified as follows:
Figure BDA00032111488700001012
wherein x (t) represents original oil pressure data, [ c ]k(t)]Represents the jth IMF component of the ith signal, and r (t) represents the residual signal.
Next, the results of the analysis by the correlation coefficient method are shown in Table 1 below.
Table 1 correlation coefficient table of IMF component and original signal
Figure BDA0003211148870000111
As can be seen from table 1, the correlation coefficients of the IMF components from 3 rd order to 8 th order to the original signal are relatively large, and are all greater than 0.1, and the IMFs 3-8 are extracted in the information entropy extraction. And then, extracting sample entropy and energy entropy of the selected IMF component in the step. Wherein, the calculation formula of the sample entropy is defined as:
Figure BDA0003211148870000112
in the formula, Bm(r)、Bm+1(r) represents the number of the maximum absolute difference of the samples of the m-dimensional and m + 1-dimensional data respectively smaller than or equal to the similarity tolerance r.
The calculation formula of the energy entropy is as follows:
Figure BDA0003211148870000113
in the formula, PiMeaning the energy of the i-th order IMF component contributes to the total energy.
In this way, since the above-obtained feature of the partial state has a certain aliasing phenomenon, the feature is fused by Kernel Principal Component Analysis (KPCA) to solve the aliasing phenomenon. The Kernel Principal Component Analysis (KPCA) is a sample nonlinear feature extraction method based on a kernel function method, and comprises the following specific calculation steps:
(1) a kernel function K is selected.
(2) Calculating a kernel function matrix:
Ki,j＝K(xi,xj)，i,j＝1,2,…,n (2-10)
(3) carrying out normalization processing on the kernel matrix:
Figure BDA0003211148870000121
(4) calculating the eigenvalue lambda of the kernel matrix KiAnd a feature vector alpha.
[λ,γ]＝eig(K)，λ＝(λ1,λ2,…,λn)，α＝[α1,α2,…,αn] (2-12)
(5) Calculating the contribution rate corresponding to the characteristic value, and then normalizing:
Figure BDA0003211148870000122
(6) according to the requirement, the original signal x is processediProjecting the k principal components to a feature vector to extract k principal components:
Figure BDA0003211148870000123
(7) the projected matrix is:
Figure BDA0003211148870000124
in the formula (I), the compound is shown in the specification,
Figure BDA0003211148870000125
which represents the k-th principal component,
Figure BDA0003211148870000126
representing the projected matrix.
The feature fusion of the obtained information entropy is realized, and a turnout fault diagnosis feature set is constructed.
Step three: and (3) improving the construction of a turnout fault diagnosis model of the SVM.
Referring to fig. 6, a turnout fault diagnosis feature set is constructed through the feature data extracted in the second step, and the feature data set is divided into a training set and a test set according to the ratio of 7: 3. Training an original SVM model by using training set data, and optimizing two parameter values of a penalty factor c and a kernel function parameter g in the SVM model by using a Tianniu Beard Algorithm (BAS) and a differential evolution algorithm (DE) respectively on the basis of the original model, so that the optimized model achieves higher fault diagnosis accuracy on a test set. Wherein, the BAS-SVM algorithm flow chart is shown in figure 7; the algorithm flow chart of the DE-SVM is shown in figure 8.
The intelligent turnout fault diagnosis method is applied to the fault diagnosis embodiment of turnout conversion equipment for verification.
Particularly, the BAS-SVM and DE-SVM models obtained by the method are selected from fault0、fault1、fault2And fault3And selecting 30 groups of samples from the four state characteristic values, and sequentially inputting the total of 30 multiplied by 4 to 120 groups of samples into the two models to carry out fault diagnosis. And 10 tests were performed for each model in order to make the diagnosis reliable. And comparing the traditional SVM model with a Relevance Vector Machine (RVM) model to obtain the statistics of the fault diagnosis rate, which is shown in the following table 2.
TABLE 2 comparison of model diagnosis rate results
Figure BDA0003211148870000131
As can be seen from the table 2, the BAS-SVM and DE-SVM models obtained by the method have turnout fault diagnosis rate results based on the fusion characteristics of more than 90%, the accuracy is higher than that of the single-characteristic turnout fault diagnosis result, and the model is remarkably improved compared with the traditional SVM model and the RVM model of the related vector machine.
The method can effectively decompose oil pressure signal data of the electro-hydraulic switch machine and extract signal characteristics aiming at fault diagnosis of the high-speed railway electro-hydraulic switch machine, and further adopts an SVM model optimized by intelligent optimization algorithm parameters to identify the fault mode.
When the intelligent turnout fault diagnosis method with the multi-feature fusion is operated through computer equipment, the computer equipment comprises the following steps: a processor (CPU) and a memory connected thereto via a bus. The processor (CPU) may execute various appropriate actions and processes according to the computer program of the above-described multi-feature integrated switch intelligent fault diagnosis method stored in a Read Only Memory (ROM) or the computer program of the above-described multi-feature integrated switch intelligent fault diagnosis method loaded into a Random Access Memory (RAM) to implement the diagnosis of the fault of the high-speed railway switch machine equipment that the above-described method can implement.
Of course, the memory also stores various programs and data necessary for the operation of the system. The computer device also includes various other components necessary for the operation of the computer.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the present invention in any way, and it will be apparent to those skilled in the art that the above description of the present invention can be applied to various modifications, equivalent variations or modifications without departing from the spirit and scope of the present invention.

Claims (10)

1. A multi-feature fused intelligent turnout fault diagnosis method is characterized by comprising the following steps:
(1) collecting oil pressure signal data of a turnout electro-hydraulic switch machine;
(2) decomposing the collected oil pressure signal data by using a MEEMD method to obtain a series of intrinsic mode functions IMF with different time scales, introducing a correlation coefficient method, screening IMF components which truly reflect the original oil pressure signals, calculating sample entropy and energy entropy, performing feature fusion on the obtained information entropy by using a kernel principal component analysis method, and constructing a turnout fault diagnosis feature data set by using the obtained result;
(3) optimizing a Support Vector Machine (SVM) by adopting a Tianniu beard search algorithm BAS and a differential evolution algorithm DE in combination with the constructed turnout fault diagnosis feature data set sample, and establishing an improved SVM turnout fault diagnosis model;
(4) and identifying and classifying different running states of the turnout by adopting the constructed improved SVM turnout fault diagnosis model to obtain a turnout fault diagnosis result.
2. The intelligent fault diagnosis method for turnout junction with multi-feature fusion as claimed in claim 1, wherein the collection of oil pressure signal data in step (1) is realized by oil pressure detection sensors arranged on left and right oil chambers of a starting oil cylinder inside the switch machine.
3. The intelligent fault diagnosis method for the multi-feature fused turnout junction according to claim 1, wherein the decomposition process of the MEEMD method on the oil pressure signal data in the step (2) is as follows:
Figure FDA0003211148860000011
wherein x (t) represents original oil pressure data, ck(t) denotes the j-th order IMF component of the i-th signal, and r (t) denotes the residual signal.
4. The intelligent fault diagnosis method for the multi-feature fused turnout junction according to claim 3, wherein the decomposition step of the MEEMD method on the oil pressure signal data in the step (2) comprises the following steps:
A1. adding two sets of white noise n with equal amplitude and standard deviation and opposite direction into original signal x (t)i(t)；
Figure FDA0003211148860000021
In the formula, aiRepresents the standard deviation of the ith white noise;
A2. decomposition by EEMD
Figure FDA0003211148860000022
And
Figure FDA0003211148860000023
obtaining Ne IMF components per noise signal;
Figure FDA0003211148860000024
Figure FDA0003211148860000025
in the formula (I), the compound is shown in the specification,
Figure FDA0003211148860000026
Figure FDA0003211148860000027
representing the jth IMF component of the ith signal.
A3. The average value of the jth IMF component of the ith signal is solved and is recorded as cj(t) subjecting
Figure FDA0003211148860000028
And
Figure FDA0003211148860000029
taking the average, i.e.
Figure FDA00032111488600000210
A4. To is calculatedTo c toj(t) EMD decomposition to obtain:
Figure FDA00032111488600000211
5. the intelligent turnout fault diagnosis method based on multi-feature fusion of claim 4, wherein in the step (2), IMF components with correlation coefficients larger than 0.1 with the original signal are screened out by introducing a correlation coefficient method, and sample entropy and energy entropy of the IMF components are calculated;
wherein, the calculation formula of the sample entropy is as follows:
Figure FDA00032111488600000212
in the formula, Bm(r)、Bm+1(r) represents the number of the maximum absolute difference of the samples of the m-dimensional and m + 1-dimensional data smaller than or equal to the similarity tolerance r, respectively;
the calculation formula of the energy entropy is as follows:
Figure FDA0003211148860000031
in the formula, PiMeaning the energy of the i-th order IMF component contributes to the total energy.
6. The intelligent turnout fault diagnosis method based on multi-feature fusion of claim 5, wherein the specific steps of performing feature fusion on the obtained information entropy by using the kernel principal component analysis method in the step (2) are as follows:
B1. selecting a kernel function K;
B2. calculating a kernel function matrix:
Ki，j＝K(xi，xj)，i，j＝1，2，…，n
B3. carrying out normalization processing on the kernel function matrix:
K′←K-InK-KIn+InKIn
B4. calculating the eigenvalue λ of the kernel function matrix KiAnd feature vector α:
[λ，γ]＝eig(K)
λ＝(λ1，λ2，…，λn)
α＝[α1，α2，…，αn]
B5. calculating the eigenvalue lambdaiCorresponding contribution rate, then normalization processing:
Figure FDA0003211148860000032
B6. the original signal xiProjecting the k principal components to a feature vector alpha to extract k principal components:
Figure FDA0003211148860000033
Figure FDA0003211148860000041
in the formula (I), the compound is shown in the specification,
Figure FDA0003211148860000042
which represents the k-th principal component,
Figure FDA0003211148860000043
representing the projected matrix.
7. The intelligent fault diagnosis method for the turnout with the multi-feature fusion according to claim 6, wherein the method for establishing the improved SVM turnout fault diagnosis model in the step (3) specifically comprises the following steps: dividing the constructed turnout fault diagnosis characteristic data set sample into a training set and a testing set according to the ratio of 7:3, training an original SVM model by using training set data, and optimizing two parameter values of a penalty factor c and a kernel function parameter g in the SVM model by using a longicorn algorithm BAS and a differential evolution algorithm DE respectively on the basis of the original SVM model so as to obtain an optimized improved SVM turnout fault diagnosis model with high fault diagnosis accuracy.
8. The intelligent fault diagnosis method for the turnout with the multi-feature fusion as claimed in claim 1, wherein the different operation states of the turnout in the step (4) comprise four states of normal operation of a switch machine, no unlocking of a turnout block, no locking of the turnout block and abnormal resistance in the conversion process.
9. A multi-feature fused switch intelligent fault diagnosis system is characterized by comprising:
the oil pressure monitoring module is used for acquiring oil pressure signal data of the turnout electro-hydraulic switch machine;
the turnout fault diagnosis characteristic data set construction module is used for decomposing and processing the collected oil pressure signal data by using an MEEMD method to obtain a series of intrinsic mode functions IMF with different time scales, introducing a correlation coefficient method, screening IMF components which truly reflect original oil pressure signals, calculating sample entropy and energy entropy of the IMF components, then performing characteristic fusion on the obtained information entropy by using a kernel principal component analysis method, and constructing a turnout fault diagnosis characteristic data set by using the obtained result;
the improved SVM turnout fault diagnosis model building module is used for optimizing a Support Vector Machine (SVM) by combining a built turnout fault diagnosis feature data set sample and adopting a Tianniu beard search algorithm BAS and a differential evolution algorithm DE to build an improved SVM turnout fault diagnosis model;
and the turnout fault diagnosis result output module is used for identifying and classifying different operation states of the turnout by adopting the constructed improved SVM turnout fault diagnosis model and outputting a turnout fault diagnosis result.
10. A computer device comprising a memory, a processor and a computer program stored in the memory and operable on the processor, wherein the processor executes the computer program to implement the multi-feature fused switch intelligent fault diagnosis method according to any one of claims 1 to 8.
CN202110930537.8A 2021-08-13 2021-08-13 Multi-feature fusion turnout intelligent fault diagnosis method, system and equipment Pending CN113627358A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110930537.8A CN113627358A (en) 2021-08-13 2021-08-13 Multi-feature fusion turnout intelligent fault diagnosis method, system and equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110930537.8A CN113627358A (en) 2021-08-13 2021-08-13 Multi-feature fusion turnout intelligent fault diagnosis method, system and equipment

Publications (1)

Publication Number Publication Date
CN113627358A true CN113627358A (en) 2021-11-09

Family

ID=78385349

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110930537.8A Pending CN113627358A (en) 2021-08-13 2021-08-13 Multi-feature fusion turnout intelligent fault diagnosis method, system and equipment

Country Status (1)

Country Link
CN (1) CN113627358A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114379619A (en) * 2021-12-09 2022-04-22 北京交通大学 Electro-hydraulic switch machine and fault diagnosis system thereof
CN115688018A (en) * 2023-01-04 2023-02-03 湖南大学 Method for monitoring state and diagnosing fault of bearing under multiple working conditions

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114379619A (en) * 2021-12-09 2022-04-22 北京交通大学 Electro-hydraulic switch machine and fault diagnosis system thereof
CN114379619B (en) * 2021-12-09 2023-05-26 北京交通大学 Electrohydraulic switch machine and fault diagnosis system thereof
CN115688018A (en) * 2023-01-04 2023-02-03 湖南大学 Method for monitoring state and diagnosing fault of bearing under multiple working conditions
CN115688018B (en) * 2023-01-04 2023-09-08 湖南大学 Method for monitoring state and diagnosing faults of bearing under multiple working conditions

Similar Documents

Publication Publication Date Title
CN106895975B (en) Bearing fault diagnosis method based on Stacked SAE deep neural network
CN108569607B (en) Elevator fault early warning method based on bidirectional gating cyclic neural network
CN113962259B (en) Multi-mode double-layer fault diagnosis method for fuel cell system
CN107657250B (en) Bearing fault detection and positioning method and detection and positioning model implementation system and method
CN108343599B (en) A kind of water pump assembly intelligent failure diagnosis method based on more granularities cascade forest
CN103291600B (en) Fault diagnosis method for hydraulic pump based on EMD-AR (empirical mode decomposition-auto-regressive) and MTS (mahalanobis taguchi system)
CN113627358A (en) Multi-feature fusion turnout intelligent fault diagnosis method, system and equipment
Bie et al. An integrated approach based on improved CEEMDAN and LSTM deep learning neural network for fault diagnosis of reciprocating pump
CN104832418A (en) Hydraulic pump fault diagnosis method based on local mean conversion and Softmax
CN115877205A (en) Intelligent fault diagnosis system and method for servo motor
CN111523659A (en) Machine fault prediction diagnosis method based on three-level neural network modeling
CN108709744A (en) Motor bearing fault diagnosis method under variable load working condition
CN105137354A (en) Motor fault detection method based on nerve network
CN110348468A (en) A kind of bearing inferior health recognition methods of the strong reconstruct edge noise reduction autocoder of Method Using Relevance Vector Machine optimization
CN115712871A (en) Power electronic system fault diagnosis method combining resampling and integrated learning
CN115958990A (en) V2G charging pile information diagnosis and safety protection method
CN113295413B (en) Traction motor bearing fault diagnosis method based on indirect signals
CN113126489B (en) Intelligent BIT design method of heavy-duty gas turbine control system based on CNN-GRU-BINN
CN115564075B (en) Main and auxiliary integrated fault collaborative diagnosis method and system for urban power grid
CN111260823B (en) Fault diagnosis method based on I/O (input/output) measuring point fault dependency matrix
CN103400021A (en) KDSNE hydroelectric generating set noise source detection method based on heavy tailed distribution
CN115456013A (en) Wind turbine generator converter power module fault diagnosis method based on operation data
CN111947954B (en) Method and system for diagnosing urban rail door system fault or sub-health
CN113139601A (en) Turnout fault detection method based on LSTM and attention mechanism
Xiao et al. Fault state identification of rolling bearings based on deep transfer convolutional autoencoder and a new health indicator

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