CN114185321A - Electric actuator fault diagnosis method for improving multi-classification twin support vector machine - Google Patents

Electric actuator fault diagnosis method for improving multi-classification twin support vector machine Download PDF

Info

Publication number
CN114185321A
CN114185321A CN202111001164.2A CN202111001164A CN114185321A CN 114185321 A CN114185321 A CN 114185321A CN 202111001164 A CN202111001164 A CN 202111001164A CN 114185321 A CN114185321 A CN 114185321A
Authority
CN
China
Prior art keywords
data
classification
fault
support vector
vector machine
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.)
Granted
Application number
CN202111001164.2A
Other languages
Chinese (zh)
Other versions
CN114185321B (en
Inventor
孙建平
李朝雅
高文捷
田乐乐
张文广
牛玉广
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.)
North China Electric Power University
Original Assignee
North China Electric Power University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by North China Electric Power University filed Critical North China Electric Power University
Priority to CN202111001164.2A priority Critical patent/CN114185321B/en
Publication of CN114185321A publication Critical patent/CN114185321A/en
Application granted granted Critical
Publication of CN114185321B publication Critical patent/CN114185321B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
    • G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
    • G05B23/0243—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B2219/00—Program-control systems
    • G05B2219/20—Pc systems
    • G05B2219/24—Pc safety
    • G05B2219/24065—Real time diagnostics

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

The invention provides a fault diagnosis method for an electric actuating mechanism of an improved multi-classification twin support vector machine, which adopts the data of the health state and the fault state of the electric actuating mechanism as an original data set; wavelet packet analysis is carried out on the data to extract characteristic vectors, and a training sample set and a test sample set are constructed. And putting the training sample set into a fault diagnosis model of the multi-twin support vector machine to obtain a classifier, and putting the test sample set into the classifier to obtain a fault diagnosis result. According to the invention, the data is subjected to feature extraction by a wavelet packet analysis method, modal aliasing during fault data of a complex actuator can be eliminated, weak data components submerged in strong data are separated, experimental errors are effectively reduced, and the fault diagnosis effect is remarkably improved.

Description

Electric actuator fault diagnosis method for improving multi-classification twin support vector machine
Technical Field
The invention belongs to the technical field of state monitoring of electric actuators of gas turbines, and particularly relates to a fault diagnosis method for an electric actuator of an improved multi-classification twin support vector machine.
Background
The electric actuator is used as an actuating mechanism of a control instruction, has a complex internal structure, is easy to break down, and has higher technical requirements on field maintenance personnel. The electric regulating valve is in various severe working environments such as high temperature, high pressure, flammability and explosiveness for a long time, and various faults are easy to occur, so that the performance and the operation level of a system are influenced, and even the economic and safety problems are caused. The basis for ensuring the safe, reliable and stable operation of an automatic control system in the industrial process is to find the fault occurring in the operation process of the regulating valve in time and take corresponding measures. At present, the electric control valve is monitored according to parameters transmitted by field equipment in real time, and generated data can be subjected to fault diagnosis and classification, so that the equipment can normally operate.
When the electric regulating valve breaks down, data can change slightly, and a large amount of environment and system noise information and the like are doped. When the traditional twin support vector machine faces small changes of data, the fault types are difficult to accurately distinguish, so that the accuracy and the reliability of data acquisition are improved, and the noise is reduced. On the basis of carrying out energy decomposition on data by wavelet packet analysis, extracting a characteristic vector, and diagnosing tiny, composite and progressive faults by adopting a method of a multi-classification twin support vector machine.
Therefore, a new method is needed for fault monitoring of various industrial actuators and diagnosing minor faults of actuators and control valves.
Disclosure of Invention
In order to solve the problems existing in the background technology, the invention provides a fault diagnosis method for an electric actuator of an improved multi-classification twin support vector machine, which is characterized by comprising the following steps:
step 1, collecting data F (t) of a gas turbine electric actuator in a healthy state and data f (t) of a fault state under different working conditions as an original data set;
step 2, extracting a characteristic vector from the original data set by utilizing wavelet packet analysis;
step 3, carrying out pathological test on the characteristic vectors, carrying out normalization processing on the characteristic vectors, constructing a training sample set and a testing sample set, and storing and recording unknown fault types;
and 4, training and testing by using a multi-classification twin support vector machine to finish fault diagnosis.
The step 2 comprises the following steps:
step 21, determining the number of wavelet packet decomposition layers and reconstructing;
step 22, extracting a data feature vector by utilizing a wavelet packet algorithm;
step 23, optimizing parameters of the multi-classification support vector machine by utilizing a wavelet packet extraction feature vector technology;
the extracting feature vectors from the wavelet packet analysis data in step 22 includes:
step 221, decomposing the coefficients of the original data set f (t), f (t): using a wavelet packet decomposition recursion method, as follows:
Figure BDA0003235689450000021
wherein, muj,n(m) is the wavelet packet coefficient of the wavelet packet function phi (t), j is the scale, uj,n(k) Is a decomposition coefficient on a scale, muj,2n(m) is an approximation of the scale j, μj,2n+1(m) is an approximation coefficient with a scale of j +1, h is a low-pass filter coefficient, g is a high-pass filter coefficient, and k is a decomposition order (k 0,1.. n);
step 222, reconstructing the coefficients of the original data sets f (t), f (t): the inverse discrete wavelet packet transform method is used as follows:
Figure BDA0003235689450000022
wherein h (m-2k) and g (m-2k) are wavelet vectors, and m is the data length;
step 223, calculating the energy of the original data sets f (t), f (t) in the frequency band, and the energy formula corresponding to each node in the third layer obtained in step 222 is as follows:
Figure BDA0003235689450000023
wherein x isjk0,1.. 7; k is 0,1.. n } is reconstruction data
Figure BDA0003235689450000024
The amplitude of the discrete points;
step 224, from step 223, constructs a feature vector as follows:
Figure BDA0003235689450000025
the data was analyzed using wavelet packets with 3 decomposition layers and 8 eigenvectors.
The step 3 comprises the following steps:
step 31, detecting whether the data exist in a fault type library by using a pathological double factor, wherein the relational expression is as follows;
Figure BDA0003235689450000026
Figure BDA0003235689450000027
wherein E is a characteristic vector, n is the number of the characteristic vectors, x is a two-dimensional function interpolation, p and q are pathological double-factor coefficients, and W is a detection parameter. When W is greater than or equal to 0, the fault type is known; when W is less than 0, the fault type is unknown;
and step 32, caching the unknown data into a fault type library, marking as fn, and marking the fault type.
After step 32, step 33 is performed to normalize the extracted feature vectors, and the relationship is as follows:
Figure BDA0003235689450000031
wherein A isijThe element representing the ith row and the jth column in the matrix, Ai'jElements in the transformed feature matrix; and dividing the normalization result of the feature vector into a training set and a test set, and inputting the training set and the test set as the feature vector into a multi-classification twin support vector machine classifier for feature recognition to perform network self-learning training and testing.
The step 4 comprises the following steps:
step 41, establishing a multi-classification twin support vector machine model; step 41 comprises: step 411 and step 412;
step 411, add lagrange multiplier alpha for i constraintsiAnd the Lagrangian function is established as follows:
Figure BDA0003235689450000032
wherein, yi(wTxi+ b) is equal to or more than 1 and is i constraint conditions, alpha is input data, b is a constraint parameter, and w isiIs a normal vector;
step 412, on the basis of step 411, constructing the ith hyperplane of the classifier:
the linear and non-linear cases of the data are first distinguished: for low-dimensional data, distinguishing the low-dimensional data in a visual mode; for high-dimensional data, distinguishing by a superposition theorem;
then, the ith hyperplane is calculated:
when the data is linear, the calculation method is as follows:
Figure BDA0003235689450000033
wherein, the obtained wiIs the normal vector of the ith hyperplane, fiIs an offset amount, giIs a penalty parameter, ξ, that takes the value as a positive numberiMore than or equal to 0 is taken as a constraint condition,
Figure BDA0003235689450000034
and
Figure BDA0003235689450000035
is a column vector with elements all being 1, Ai、BiThe ith hyperplane fitting coefficients of the two classifiers are respectively obtained;
when the data is in a nonlinear condition, the calculation method is as follows:
Figure BDA0003235689450000036
wherein u isiIs the normal vector of the ith hyperplane, fiIs an offset amount, giIs a penalty parameter, ξ, that takes the value as a positive numberiMore than or equal to 0 is taken as a constraint condition,
Figure BDA0003235689450000037
and
Figure BDA0003235689450000038
is a column vector with elements all being 1, Ai、BiThe ith hyperplane fitting coefficients of the two classifiers are respectively represented, and N, C is a kernel function fixed parameter;
step 42, obtaining a decision function Lable (x) of the multi-classification twin support vector machine according to the step 412 as follows:
when the data is linear:
Figure BDA0003235689450000041
when the data is a non-linear case:
Figure BDA0003235689450000042
and (4) taking a decision function obtained by training and testing of the multi-classification twin support vector machine as a fault diagnosis classification result, and comparing the fault diagnosis result with the initially set fault type and fault strength to judge the effect of the fault diagnosis method to be verified.
The invention has the beneficial effects that:
1. wavelet packet analysis is added into a traditional multi-classification twin support vector machine to extract feature vectors, and the recognition degree of a multi-classification twin support vector machine model is improved.
2. The method can accurately judge and classify various states, including faults with extremely small change of the data of the opening degree of the valve, such as motor faults, valve blockage and the like.
Drawings
FIG. 1 is a flow chart of a method of the present invention for improving fault diagnosis of an electric actuator of a multi-classification twin support vector machine;
FIG. 2 is a block diagram of an electric actuator of the present invention;
FIG. 3 is a wavelet packet analysis feature vector diagram used in the present invention;
FIG. 4 is a schematic diagram of a wavelet packet analysis structure used in the present invention;
FIG. 5 is a diagram of the diagnostic effect of the fault diagnosis method of the improved multi-classification twin support vector machine of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
The embodiment of the present invention shown in fig. 1 includes: the method comprises the following steps of:
step 1, collecting data F (t) of a gas turbine electric actuator in a healthy state and data f (t) of a fault state under different working conditions as an original data set;
step 2, extracting characteristic vectors from the original data sets F (t), f (t) by utilizing wavelet packet analysis;
step 3, carrying out pathological test on the feature vectors, carrying out normalization processing on the feature vectors, constructing a training sample set and a testing sample set, and storing unknown fault types into a fault type library;
and 4, training and testing by using a multi-classification twin support vector machine to finish fault diagnosis.
Wherein the step 2 is divided into:
step 21, determining the number of wavelet packet decomposition layers and reconstructing;
step 22, extracting a data feature vector by utilizing a wavelet packet algorithm;
and step 23, optimizing parameters of the multi-classification support vector machine by utilizing a wavelet packet extraction feature vector technology.
Extracting feature vectors from the wavelet packet analysis data in step 22: for coefficient decomposition, coefficient reconstruction, frequency bandwidth energy calculation and eigenvector generation of data, the specific steps are as follows:
step 221, decomposing the coefficients of the original data set f (t), f (t): using a wavelet packet decomposition recursion method, as follows:
Figure BDA0003235689450000051
wherein, muj,n(m) is the wavelet packet coefficient of the wavelet packet function phi (t), j is the scale, uj,n(k) Is a decomposition coefficient on a scale, muj,2n(m) is an approximation of the scale j, μj,2n+1(m) is an approximation coefficient with a scale j +1, h is a low-pass filter coefficient, g is a high-pass filter coefficient, and k is a decomposition order (k 0,1.. n).
Step 222, reconstructing the coefficients of the original data sets f (t), f (t): the inverse discrete wavelet packet transform method is used as follows:
Figure BDA0003235689450000052
wherein h (m-2k) and g (m-2k) are wavelet vectors, and m is the data length.
Step 223, calculating the energy of the original data sets f (t), f (t) in the frequency band, and the energy formula corresponding to each node in the third layer obtained in step 222 is as follows:
Figure BDA0003235689450000053
wherein x isjk0,1.. 7; k is 0,1.. n } is reconstruction data
Figure BDA0003235689450000054
The amplitude of the discrete points.
Step 224, from step 223, constructs a feature vector as follows:
Figure BDA0003235689450000055
further, as shown in fig. 4, the data is analyzed using wavelet packets, the number of decomposition layers is 3, and the eigenvector is 8.
Step 3 is feature vector screening, which is divided into: step 31, ill-conditioned test of the feature vector, step 32, storage and recording of unknown fault types, and step 33, normalization processing of each row of the feature vector, specifically as follows:
step 31, detecting whether the data exist in a fault type library by using a pathological double factor, wherein the relational expression is as follows;
Figure BDA0003235689450000061
Figure BDA0003235689450000062
wherein E is a characteristic vector, n is the number of the characteristic vectors, x is a two-dimensional function interpolation, p and q are pathological double-factor coefficients, and W is a detection parameter. When W is greater than or equal to 0, the fault type is known; when W is less than 0, the fault type is unknown.
And step 32, caching unknown data into a fault type library by using a setlocal calling program in matlab, recording the unknown data as fn, and marking the fault type so as to call the unknown data in the later fault diagnosis, wherein the step is shown in table 1.
TABLE 1 failure type library
Fault number Name of failure
F1 Valve plug
F2 Valve or valve seat sinking
F3 Valve plug or seat erosion
F4 Increase of bearing friction
F5 External leakage
F6 Internal leakage
F7 Over-voltage
F8 Actuator push rod distortion
F9 Loosening of frames or racks
F10 Spring failure
F11 Electrical converter failure
F12 Valve stem displacement sensor failure
F13 Pressure sensor failure
F14 Positioner feedback fault
F15 Electrical bypass valve failure
Step 33, normalizing each line of the feature vectors extracted by wavelet packet analysis, wherein the relation is as follows:
Figure BDA0003235689450000063
wherein A isijThe element representing the ith row and the jth column in the matrix, Ai'jAre elements in the transformed feature matrix. And dividing the normalization result of the feature vector into a training set and a test set, and inputting the training set and the test set as the feature vector into a multi-classification twin support vector machine classifier for feature recognition to perform network self-learning training and testing.
Wherein the step 4 is divided into:
performing multi-classification twin support vector machine training, comprising: step 41 of establishing a multi-classification twin support vector machine model and step 42 of obtaining a fault diagnosis classification result through a decision function, which are specifically as follows:
and 41, establishing a multi-classification twin support vector machine model. The specific method comprises the following steps:
step 411, add lagrange multiplier alpha for i constraintsiAnd the Lagrangian function is established as follows:
Figure BDA0003235689450000071
wherein, yi(wTxi+ b) is equal to or more than 1 and is i constraint conditions, alpha is input data, b is a constraint parameter, and w isiIs a normal vector.
Step 412, on the basis of step 411, constructing the ith hyperplane of the classifier:
the linear and non-linear cases of the data are first distinguished: for low-dimensional data, distinguishing the low-dimensional data in a visual mode; for high-dimensional data, distinguishing by a superposition theorem;
then, the ith hyperplane is calculated:
when the data is linear, the calculation method is as follows:
Figure BDA0003235689450000072
wherein, the obtained wiIs the normal vector of the ith hyperplane, fiIs an offset amount, giIs a penalty parameter, ξ, that takes the value as a positive numberiMore than or equal to 0 is taken as a constraint condition,
Figure BDA0003235689450000073
and
Figure BDA0003235689450000074
is a column vector with elements all being 1, Ai、BiThe ith hyperplane fitting coefficients of the two classifiers are respectively obtained.
When the data is in a nonlinear condition, the calculation method is as follows:
Figure BDA0003235689450000075
wherein u isiIs the normal vector of the ith hyperplane, fiIs an offset amount, giIs a penalty parameter which takes the value as a positive number，ξiMore than or equal to 0 is taken as a constraint condition,
Figure BDA0003235689450000076
and
Figure BDA0003235689450000077
is a column vector with elements all being 1, Ai、BiThe i-th hyperplane fit coefficients of the two classifiers are respectively, and N, C is a kernel function fixed parameter.
Step 42, obtaining a decision function Lable (x) of the multi-classification twin support vector machine according to the step 412 as follows:
when the data is linear:
Figure BDA0003235689450000081
when the data is a non-linear case:
Figure BDA0003235689450000082
and a decision function Lable (x) obtained by training and testing of the multi-classification twin support vector machine is used as a fault diagnosis classification result, and the fault diagnosis result is compared with the initially set fault type and fault strength to judge the effect of the fault diagnosis method to be verified.
Aiming at the defects of data analysis of the traditional multi-classification twin support vector machine, the invention optimizes decision parameters of the multi-classification twin support vector machine by utilizing wavelet packet analysis and extracts the characteristics of the data of the electric actuator. And the unknown fault types are stored and recorded, then the characteristic vectors are processed in a normalized mode to form a training set and a testing set of the multi-classification twin support vector machine, and the decision function parameters can be automatically adjusted to diagnose and classify the data.
According to the diagnosis method for the improved multi-classification twin support vector machine, provided by the invention, through wavelet packet analysis, modal aliasing during complex actuator fault data can be eliminated, weak data components submerged in strong data can be separated out, and experimental errors are effectively reduced. Meanwhile, by means of the pathological test, when the fault data are judged to be unknown, the new data are recorded into the fault type library for subsequent use, and numerous problems caused by the fact that the unknown data cannot be accurately diagnosed are solved. And the multi-classification twin support vector machine can rapidly diagnose a plurality of faults at one time, has small error and high convergence speed when executing a fault classification task, and can more efficiently complete the classification task.
Table 1 shows the comparison of the results of the fault diagnosis of the method of the present invention with other methods, which is innovative in that the fault can be quickly diagnosed, and based on the same set of data, the method is used only for 1.708s, and the diagnosis speed is improved by nearly 6 times compared with the conventional method. The highest accuracy of the method is 94.341%, and compared with the traditional method, the diagnosis accuracy is improved by nearly 10%. When unknown faults exist, the method can accurately strip the unknown faults, fault error diagnosis is avoided, the defect that the traditional diagnosis method cannot completely classify the unknown faults is overcome, and the fault diagnosis effect when fault data contain the unknown faults is greatly improved.
The results of the present invention are shown in Table 2 for comparison with the results of the conventional methods for accuracy and time.
TABLE 1 comparison of the results of the present invention with those of the conventional methods
Algorithm Rate of accuracy/%) Time/s
SVM 83.051 10.235
WPT-SVM 87.232 9.753
TWSVM 90.702 7.924
WPT-TWSVM 91.250 7.285
OVA-TWSVM 90.980 2.008
WPT-OVA-TWSVM 94.341 1.708

Claims (5)

1. An electric actuator fault diagnosis method for improving a multi-classification twin support vector machine is characterized by comprising the following steps:
step 1, collecting data F (t) of a gas turbine electric actuator in a healthy state and data f (t) of a fault state under different working conditions as an original data set;
step 2, extracting a characteristic vector from the original data set by utilizing wavelet packet analysis;
step 3, carrying out pathological test on the characteristic vectors, carrying out normalization processing on the characteristic vectors, constructing a training sample set and a testing sample set, and storing and recording unknown fault types;
and 4, training and testing by using a multi-classification twin support vector machine to finish fault diagnosis.
2. The method for improving the fault diagnosis of the electric actuator of the multi-classification twin support vector machine according to claim 1, wherein the step 2 comprises the following steps:
step 21, determining the number of wavelet packet decomposition layers and reconstructing;
step 22, extracting a data feature vector by utilizing a wavelet packet algorithm;
step 23, optimizing parameters of the multi-classification support vector machine by utilizing a wavelet packet extraction feature vector technology;
the extracting feature vectors from the wavelet packet analysis data in step 22 includes:
step 221, decomposing the coefficients of the original data set f (t), f (t): using a wavelet packet decomposition recursion method, as follows:
Figure FDA0003235689440000011
wherein, muj,n(m) is the wavelet packet coefficient of the wavelet packet function phi (t), j is the scale, uj,n(k) Is a decomposition coefficient on a scale, muj,2n(m) is an approximation of the scale j, μj,2n+1(m) is an approximation coefficient with a scale of j +1, h is a low-pass filter coefficient, g is a high-pass filter coefficient, and k is a decomposition order (k 0,1.. n);
step 222, reconstructing the coefficients of the original data sets f (t), f (t): the inverse discrete wavelet packet transform method is used as follows:
Figure FDA0003235689440000012
wherein h (m-2k) and g (m-2k) are wavelet vectors, and m is the data length;
step 223, calculating the energy of the original data sets f (t), f (t) in the frequency band, and the energy formula corresponding to each node in the third layer obtained in step 222 is as follows:
Figure FDA0003235689440000013
wherein x isjk0,1.. 7; k is 0,1.. n } is reconstruction data
Figure FDA0003235689440000015
The amplitude of the discrete points;
step 224, from step 223, constructs a feature vector as follows:
Figure FDA0003235689440000014
the data was analyzed using wavelet packets with 3 decomposition layers and 8 eigenvectors.
3. The method for improving the fault diagnosis of the electric actuator of the multi-classification twin support vector machine according to claim 1, wherein the step 3 comprises the following steps:
step 31, detecting whether the data exist in a fault type library by using a pathological double factor, wherein the relational expression is as follows;
Figure FDA0003235689440000021
Figure FDA0003235689440000022
wherein E is a characteristic vector, n is the number of the characteristic vectors, x is a two-dimensional function interpolation, p and q are pathological double-factor coefficients, and W is a detection parameter. When W is greater than or equal to 0, the fault type is known; when W is less than 0, the fault type is unknown;
and step 32, caching the unknown data into a fault type library, marking as fn, and marking the fault type.
4. The method for diagnosing the fault of the electric actuator of the improved multi-classification twin support vector machine as claimed in claim 3, wherein after the step 32, the step 33 of normalizing the extracted feature vectors is performed for each row, and the relation is as follows:
Figure FDA0003235689440000023
wherein A isijRepresents the element of the ith row and jth column in the matrix, A'ijElements in the transformed feature matrix; and dividing the normalization result of the feature vector into a training set and a test set, and inputting the training set and the test set as the feature vector into a multi-classification twin support vector machine classifier for feature recognition to perform network self-learning training and testing.
5. The method for improving the fault diagnosis of the electric actuator of the multi-classification twin support vector machine according to claim 1, wherein the step 4 comprises the following steps:
step 41, establishing a multi-classification twin support vector machine model; step 41 comprises: step 411 and step 412;
step 411, add lagrange multiplier alpha for i constraintsiAnd the Lagrangian function is established as follows:
Figure FDA0003235689440000024
wherein, yi(wTxi+ b) is equal to or more than 1 and is i constraint conditions, alpha is input data, b is a constraint parameter, and w isiIs a normal vector;
step 412, on the basis of step 411, constructing the ith hyperplane of the classifier:
the linear and non-linear cases of the data are first distinguished: for low-dimensional data, distinguishing the low-dimensional data in a visual mode; for high-dimensional data, distinguishing by a superposition theorem;
then, the ith hyperplane is calculated:
when the data is linear, the calculation method is as follows:
Figure FDA0003235689440000031
wherein, the obtained wiIs the normal vector of the ith hyperplane, fiIs an offset amount, giIs a penalty parameter, ξ, that takes the value as a positive numberiMore than or equal to 0 is taken as a constraint condition,
Figure FDA0003235689440000032
and
Figure FDA0003235689440000033
is a column vector with elements all being 1, Ai、BiThe ith hyperplane fitting coefficients of the two classifiers are respectively obtained;
when the data is in a nonlinear condition, the calculation method is as follows:
Figure FDA0003235689440000034
wherein u isiIs the normal vector of the ith hyperplane, fiIs an offset amount, giIs a penalty parameter, ξ, that takes the value as a positive numberiMore than or equal to 0 is taken as a constraint condition,
Figure FDA0003235689440000035
and
Figure FDA0003235689440000036
is a column vector with elements all being 1, Ai、BiThe ith hyperplane fitting coefficients of the two classifiers are respectively represented, and N, C is a kernel function fixed parameter;
step 42, obtaining a decision function Lable (x) of the multi-classification twin support vector machine according to the step 412 as follows:
when the data is linear:
Figure FDA0003235689440000037
when the data is a non-linear case:
Figure FDA0003235689440000038
and (4) taking a decision function obtained by training and testing of the multi-classification twin support vector machine as a fault diagnosis classification result, and comparing the fault diagnosis result with the initially set fault type and fault strength to judge the effect of the fault diagnosis method to be verified.
CN202111001164.2A 2021-08-30 2021-08-30 Electric actuator fault diagnosis method for improving multi-classification twin support vector machine Active CN114185321B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111001164.2A CN114185321B (en) 2021-08-30 2021-08-30 Electric actuator fault diagnosis method for improving multi-classification twin support vector machine

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111001164.2A CN114185321B (en) 2021-08-30 2021-08-30 Electric actuator fault diagnosis method for improving multi-classification twin support vector machine

Publications (2)

Publication Number Publication Date
CN114185321A true CN114185321A (en) 2022-03-15
CN114185321B CN114185321B (en) 2023-12-22

Family

ID=80601043

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111001164.2A Active CN114185321B (en) 2021-08-30 2021-08-30 Electric actuator fault diagnosis method for improving multi-classification twin support vector machine

Country Status (1)

Country Link
CN (1) CN114185321B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102539159A (en) * 2010-12-24 2012-07-04 中国船舶研究设计中心 Fault diagnosis method for valve mechanism of diesel engine
CN107101813A (en) * 2017-04-26 2017-08-29 河北工业大学 A kind of frame-type circuit breaker mechanical breakdown degree assessment method based on vibration signal
US20180144246A1 (en) * 2016-11-16 2018-05-24 Indian Institute Of Technology Delhi Neural Network Classifier
CN108537260A (en) * 2018-03-29 2018-09-14 上海交通大学 A kind of crane transmission axis method for diagnosing faults and system

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102539159A (en) * 2010-12-24 2012-07-04 中国船舶研究设计中心 Fault diagnosis method for valve mechanism of diesel engine
US20180144246A1 (en) * 2016-11-16 2018-05-24 Indian Institute Of Technology Delhi Neural Network Classifier
CN107101813A (en) * 2017-04-26 2017-08-29 河北工业大学 A kind of frame-type circuit breaker mechanical breakdown degree assessment method based on vibration signal
CN108537260A (en) * 2018-03-29 2018-09-14 上海交通大学 A kind of crane transmission axis method for diagnosing faults and system

Non-Patent Citations (7)

* Cited by examiner, † Cited by third party
Title
GUO XIAOXUAN: "EEMD Method and TWSVM for Fault Diagnosis of Roller Bearings", 《PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON MANAGEMENT, COMPUTER AND EDUCATION INFORMATIZATION》, vol. 25, pages 102 - 106 *
ZHAOYA LI: "Research on Fault Diagnosis of Electric Control Valve Based on WPT-TWSVM", 《2021 36TH YOUTH ACADEMIC ANNUAL CONFERENCE OF CHINESE ASSOCIATION OF AUTOMATION (YAC)》, pages 649 - 654 *
ZHIWEN LIU: "A hybrid intelligent multi-fault detection method for rotating machinery based on RSGWPT, KPCA and Twin SVM", 《ISA TRANSACTIONS》, vol. 66, pages 249 - 261, XP029903624, DOI: 10.1016/j.isatra.2016.11.001 *
刘军科: "基于孪生支持向量机的齿轮箱故障诊断", 《自动化技术与应用》, vol. 39, no. 7, pages 5 - 10 *
刘军科: "基于支持向量机的风机齿轮箱故障诊断的研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》, no. 1, pages 042 - 454 *
徐冠基: "基于SPSO优化Multiple Kernel-TWSVM的滚动轴承故障诊断", 《振动.测试与诊断》, vol. 39, no. 5, pages 973 - 979 *
许小刚: "基于小波包能量分析及改进支持向量机的风机机械故障诊断", 《动 力 工 程 学 报》, vol. 33, no. 8, pages 606 - 612 *

Also Published As

Publication number Publication date
CN114185321B (en) 2023-12-22

Similar Documents

Publication Publication Date Title
CN109522600B (en) Complex equipment residual service life prediction method based on combined deep neural network
CN109446187B (en) Method for monitoring health state of complex equipment based on attention mechanism and neural network
CN110071913B (en) Unsupervised learning-based time series anomaly detection method
CN109146246B (en) Fault detection method based on automatic encoder and Bayesian network
CN111780800B (en) Method and system for monitoring, isolating and reconstructing sensor fault
CN112257530B (en) Rolling bearing fault diagnosis method based on blind signal separation and support vector machine
CN113642754B (en) Complex industrial process fault prediction method based on RF noise reduction self-coding information reconstruction and time convolution network
CN111523081B (en) Aeroengine fault diagnosis method based on enhanced gate control circulating neural network
CN116226646B (en) Method, system, equipment and medium for predicting health state and residual life of bearing
CN109710983B (en) Diesel engine cylinder layered fault diagnosis method based on key performance indexes
CN113537328B (en) Rotary machine fault diagnosis method and device based on deep learning
CN114239377A (en) Method and system for evaluating health state of urban rail electromechanical equipment and storage medium
CN113657454B (en) Nuclear power rotating machinery state monitoring method based on autoregressive BiGRU
CN111103137A (en) Wind turbine gearbox fault diagnosis method based on deep neural 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
CN114048688A (en) Method for predicting service life of bearing of wind power generator
CN114462508A (en) Power transformer health state assessment method based on multi-mode neural network
CN114936520B (en) Steam turbine fault early warning analysis method based on improved MSET
CN115204272A (en) Industrial system fault diagnosis method and equipment based on multi-sampling rate data
CN113468728B (en) Variable pitch system fault prediction method based on neural network
CN114741945A (en) Diagnosis method for online faults of aircraft engine
CN113703422A (en) Gas turbine pneumatic actuator fault diagnosis method based on characteristic analysis processing
CN114185321B (en) Electric actuator fault diagnosis method for improving multi-classification twin support vector machine
Jianhui et al. New fault diagnosis method for rolling bearing based on PCA
CN105279553B (en) A kind of height adds to water system fault degree recognition methods

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