CN111091159A - Weak fault feature extraction method for autonomous underwater robot propeller - Google Patents

Weak fault feature extraction method for autonomous underwater robot propeller Download PDF

Info

Publication number
CN111091159A
CN111091159A CN201911372169.9A CN201911372169A CN111091159A CN 111091159 A CN111091159 A CN 111091159A CN 201911372169 A CN201911372169 A CN 201911372169A CN 111091159 A CN111091159 A CN 111091159A
Authority
CN
China
Prior art keywords
fault
atom
decomposition
sparse decomposition
weight matrix
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
CN201911372169.9A
Other languages
Chinese (zh)
Other versions
CN111091159B (en
Inventor
张铭钧
张天驰
陈泽宇
于大程
吕图
崔丁与
盖宁
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.)
Harbin Engineering University
Original Assignee
Harbin Engineering 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 Harbin Engineering University filed Critical Harbin Engineering University
Priority to CN201911372169.9A priority Critical patent/CN111091159B/en
Publication of CN111091159A publication Critical patent/CN111091159A/en
Application granted granted Critical
Publication of CN111091159B publication Critical patent/CN111091159B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • G06F18/2136—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on sparsity criteria, e.g. with an overcomplete basis
    • 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)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Testing And Monitoring For Control Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a weak fault feature extraction method for a propeller of an autonomous underwater robot, and belongs to the technical field of fault diagnosis of underwater robots. The method firstly adds a time variable to the traditional sparse decomposition algorithm. Each beat of the signal to be decomposed can obtain a corresponding decomposition result, so that the precision of the decomposition result in a time domain is higher; then enhancing fault characteristics on the basis of a traditional characteristic extraction method, and storing fault information in a fault weight matrix by establishing the fault weight matrix corresponding to the atomic dictionary; and finally, taking the inner product of the sparse decomposition coefficient and the fault weight matrix as the enhanced fault characteristic, and further increasing the difference value between the fault characteristic and the interference characteristic. The method can effectively highlight and extract fault characteristics, is convenient for subsequent detection and identification of weak faults of the propeller, and is particularly suitable for monitoring the state of the propeller of the autonomous underwater robot.

Description

Weak fault feature extraction method for autonomous underwater robot propeller
Technical Field
The invention relates to a weak fault feature extraction method of an autonomous underwater robot propeller based on combination of sparse decomposition and a fault weight matrix, and belongs to the technical field of underwater robot fault diagnosis.
Background
An Autonomous Underwater Vehicle (AUV) unmanned cable works in the marine environment, and the safety is one of important research contents in the research and practical process of the AUV. The propeller is the most important power component in the AUV, and is the component which usually fails, so that if the failure can be diagnosed at the early stage of the failure, the more serious failure can be avoided. The faults of the propeller are weak faults (the thrust loss fault degree of the propeller is less than 10%), so that the method for diagnosing the weak faults of the propeller has important research significance.
Compared with the problem of fault feature extraction in other fields, the AUV propeller weak fault feature extraction has the advantages that the AUV works in the water environment, so that the whole AUV system is influenced by random ocean current interference, and the fault feature extraction result is influenced. Finding out that the AUV propeller fault characteristics are extracted based on a typical wavelet decomposition method: the method has good diagnosis effect on the faults with large propeller output loss (more than 35%); however, for a fault with a small propeller output loss (less than 10%), wavelet detail coefficients serving as fault features after wavelet decomposition are hidden in interference features, and abnormal points of the fault are difficult to distinguish through the wavelet detail coefficients, so that misdiagnosis or missed diagnosis can occur during final diagnosis.
Through theory and relevant data analysis, the method has the following findings: due to the short signal change time caused by the fault, when wavelet decomposition is used, a low decomposition scale with a small time window width needs to be selected for analysis. But because the frequency band is wider at the low scale of wavelet decomposition, and the frequency band of the fault feature and the frequency band of the interference have aliasing to some extent. Therefore, the wavelet detail coefficients as the fault features have higher amplitude of the fault features and higher amplitude of the interference features in the fault feature frequency band. Thus leading to the advance being carried out
Disclosure of Invention
The invention aims to provide a weak fault feature extraction method of an autonomous underwater robot propeller based on combination of sparse decomposition and a fault weight matrix. The method can effectively highlight and conveniently extract the failure characteristics of the propeller so as to facilitate the subsequent failure detection and identification work and ensure the safe and reliable operation of the AUV.
The purpose of the invention is realized as follows: the method comprises the following steps:
step 1: performing atomic dictionary learning, namely performing dictionary learning by using an existing data sample to obtain a basis function similar to the structural characteristics of the AUV state signal;
step 2: sparse decomposition, which is to perform sparse decomposition on the signal based on the learnt atom dictionary and the time variable;
and step 3: continuously carrying out sparse decomposition on the fault sample, and updating a fault weight matrix;
and 4, step 4: and performing inner product operation on the sparse decomposition coefficients corresponding to the atom dictionaries obtained after sparse decomposition and the obtained fault weight matrix, and taking the calculation result as a fault characteristic result.
The invention also includes such structural features:
1. the step 2 specifically comprises the following steps:
the set of the atom dictionary at any time tau is obtained by adding a time shift operator, and the expression of the k-th iteration sparse decomposition is as follows:
Figure BDA0002339965110000021
wherein:
Figure BDA0002339965110000022
representing the input signal, TτRepresents the time shift operator, mkRepresenting a collection of atom dictionaries, TτmkI.e. represented as a collection of atom dictionaries at each time instant other than the time instant t, αk,τRepresenting the sparse decomposition coefficients, epsilon represents the sparse decomposition residual terms.
The corresponding constraints are:
Figure BDA0002339965110000023
in the formula: l0The norm represents the number of non-zero terms in the coefficient vector, is a non-deterministic polynomial, namely cannot be directly solved, and is generally selected from l in practical application2Norm is approximated, so the above equation is further represented by l2The norm is expressed as:
Figure BDA0002339965110000024
the equation for obtaining the best-matched atom and the corresponding sparse decomposition coefficient obtained by the kth iteration in the decomposition process based on the above formula is:
Figure BDA0002339965110000025
the above formula represents the most matched atom in the k-th iteration and the constraint condition met by the corresponding sparse decomposition coefficient, and the most matched atom m is finally obtained by solving the above formulakAnd corresponding sparse decomposition coefficients αk,τThe expression of (a) is:
Figure BDA0002339965110000026
Figure BDA0002339965110000027
wherein,
Figure BDA0002339965110000028
representing the time-shift operator Tταk,τRepresents the best matching atom m in the final decomposition resultkCorresponding sparse decomposition coefficients.
2. The step 3 specifically comprises:
1) initialization
Taking J pieces to be treatedLearning sample { x1，x2…xJJ is 1, and the residual error term is initialized to be epsilon and xjThe atom dictionary D is an atom dictionary which is learned based on a K-SVD algorithm, the iteration time t is 1, and a coefficient matrix is decomposed
Figure BDA0002339965110000031
Fault weight matrix
Figure BDA0002339965110000032
2) Finding matching atoms
Calculating residual error epsilon and each atom D in atom dictionary DjAnd find the lower subscript λ of the maximum value of the inner product, i.e.:
|<x,dj>|＝supi∈(1,…,k)|<x,dλ>|
3) updating coefficient matrix and fault weight matrix
The decomposition coefficient matrix is updated as:
αt＝αt-1∪|<x,dλ>|
and updating the corresponding fault weight matrix as follows:
Wλ＝Wλ+αt
4) updating residual errors
The residual is the difference between the original signal and the current iteration fitting result, that is:
εt＝x-Dtαt
5) determining termination condition
The iteration termination condition of the algorithm is the number i of atoms used for decomposing the signal, and if the iteration times t is less than i, the step 1) is returned; and if the iteration time t is more than or equal to i, learning the next group of data, namely j is j +1, carrying out normalization processing on the fault weight matrix when all data are completely learned, and then stopping the algorithm.
Compared with the prior art, the invention has the beneficial effects that: compared with the prior art, the invention has the following beneficial effects: the reason that the time domain precision is poor when the traditional sparse decomposition algorithm decomposes the time sequence signal is that the traditional method cannot directly decompose the long signal, but needs to divide the signal and then decompose the signal respectively. For this reason, the patent improves the traditional sparse decomposition algorithm. Adding the time shift operator into the calculation formula of the traditional sparse decomposition algorithm is equivalent to adding a time variable to the traditional sparse decomposition algorithm. Each beat of the signal to be decomposed will get a corresponding decomposition result, instead of the same decomposition result in the same time slice of the conventional method. Thereby making the decomposition result more accurate in the time domain. In addition, in the traditional method, only sparse decomposition coefficients are used as fault characteristics, and weak faults of the propeller and ocean current interference are difficult to distinguish. According to the method, fault characteristics are enhanced on the basis of a traditional characteristic extraction method, fault information is stored in a fault weight matrix by establishing the fault weight matrix corresponding to an atomic dictionary, and then the inner product of a sparse decomposition coefficient and the fault weight matrix is used as the enhanced fault characteristics. Thereby increasing the difference between the fault signature and the interference signature. The invention can effectively highlight and conveniently extract the failure characteristics of the propeller so as to facilitate the subsequent failure detection and identification work and ensure the safe and reliable operation of the AUV.
Drawings
Fig. 1 is a flowchart of a method for extracting a fault feature according to the present invention.
FIG. 2 is a flowchart of a method for learning a fault weight matrix according to the present invention.
FIG. 3 is the comparison result of the longitudinal speed fault characteristics of the invention.
FIG. 4 shows the comparison result of the heading fault characteristics of the invention.
FIG. 5 is a comparison result of main-boost voltage fault characteristics of the present invention.
FIG. 6 is a comparison result of the effect of the method for extracting the fault characteristics of the patent of the present invention
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
Fig. 1 is a flow chart of the method for extracting the AUV fault characteristics of the present invention. With reference to fig. 1, the weak fault feature extraction method of the autonomous underwater robot propeller based on the combination of sparse decomposition and the fault weight matrix includes the following specific implementation steps:
step 1: based on the existing dictionary learning method, the existing data samples are utilized to carry out dictionary learning, and basis functions similar to the structural features of the AUV state signals are obtained.
Step 2: and performing sparse decomposition on the signal based on the learnt atom dictionary and the time variable.
Because the improved algorithm of the invention is to decompose a long signal (with the length of K) by an atom dictionary with a fixed length, the invention adds a time shift operator on the basis of the existing sparse decomposition method to obtain a set of the atom dictionary at any moment, and the K-th iteration sparse decomposition expression is as follows:
Figure BDA0002339965110000041
wherein:
Figure BDA0002339965110000042
representing the input signal, TτRepresents the time shift operator, mkRepresenting a collection of atom dictionaries, TτmkI.e. represented as the set of atomic dictionaries at each time other than time τ αk,τRepresenting a sparse decomposition coefficient, and epsilon represents a sparse decomposition residual error term;
the corresponding constraint can be expressed as:
Figure BDA0002339965110000043
formulas (1) and (2) are calculation formulas of the improved sparse decomposition algorithm of the patent. However, the formulas (1) and (2) are ideal expressions of a sparse component solution, and a decomposition result to be finally obtained in practical application is an atom dictionary set mkCorresponding sparse decomposition coefficients αk,τSo that the decomposition coefficient α needs to be further derived by the improved formulas (1) and (2) of this patentk,τThe calculation formula of (2).
The norm in the formula (2) represents the number of non-zero terms in the coefficient vector, and is a non-deterministic polynomial, that is, the non-deterministic polynomial cannot be directly solved, and generally is represented by a norm to approximate in practical application, so the above formula is further represented by a norm:
Figure BDA0002339965110000044
based on formula (3), the best matching atom and corresponding sparse decomposition coefficient obtained in the k-th iteration in the decomposition process can be obtained as follows:
Figure BDA0002339965110000051
formula (4) represents the most matched atom and the constraint condition satisfied by the corresponding sparse decomposition coefficient in the k-th iteration, and the final most matched atom m is obtained by solving formula (4)kAnd corresponding sparse decomposition coefficients αk,τThe expression of (a) is:
Figure BDA0002339965110000052
Figure BDA0002339965110000053
in the formulas (5) and (6)
Figure BDA0002339965110000054
Representing the time-shift operator Tταk,τRepresents the best matching atom m in the final decomposition resultkCorresponding sparse decomposition coefficients.
And step 3: and continuously carrying out sparse decomposition on the fault samples, and updating the fault weight matrix.
The specific idea of the fault weight matrix learning method is as follows: the atomic dictionary of sparse decomposition belongs to a label-free dictionary, and does not contain the fault state information required by the text. Therefore, when the atom dictionary is built, the fault label, namely the fault weight matrix mentioned in the patent, is added to each atom, so that the effect of feature enhancement is achieved. Therefore, the method decomposes a large number of known fault data samples, the decomposition coefficients contain required fault information, and then the decomposed decomposition coefficients of the same atoms are accumulated to form the fault weight matrix required by the method.
The learning method of the fault weight matrix can be divided into the following five steps, as shown in fig. 2, the flow is as follows:
1) initialization
Taking J samples to be learned { x1，x2…xJJ is 1, and the residual error term is initialized to be epsilon and xjThe atom dictionary D is an atom dictionary which is learned based on a K-SVD algorithm, the iteration time t is 1, and a coefficient matrix is decomposed
Figure BDA0002339965110000055
Fault weight matrix
Figure BDA0002339965110000056
2) Finding matching atoms
Calculating residual error epsilon and each atom D in atom dictionary DjAnd find the lower subscript λ of the maximum value of the inner product, i.e.:
|<x,dj>|＝supi∈(1,…,k)|<x,dλ>| (7)
3) updating coefficient matrix and fault weight matrix
The decomposition coefficient matrix is updated as:
αt＝αt-1∪|<x,dλ>| (8)
and updating the corresponding fault weight matrix as follows:
Wλ＝Wλ+αt(9)
4) updating residual errors
The residual is the difference between the original signal and the current iteration fitting result, that is:
εt＝x-Dtαt(10)
5) determining termination condition
The iteration termination condition of the algorithm is the number i of atoms used for decomposing the signal. If the iteration times t < i, returning to the step 1); and if the iteration time t is more than or equal to i, learning the next group of data, namely j is j +1, carrying out normalization processing on the fault weight matrix when all data are completely learned, and then stopping the algorithm. Where i is the number of atoms used to decompose the signal, and because of the sparseness of the signal, only a few atoms contain useful information about the signal. The selection method can analyze and select the decomposition coefficient amplitude values corresponding to different atoms after the fault sample is decomposed. Taking the research in this document as an example, the research in this document finds that the decomposition coefficient corresponding to 2-3 atoms in the signal decomposition result of the AUV sensor is usually much larger than the rest atoms, so the coefficient i is set to 3 in the algorithm in this document.
And 4, step 4: and performing inner product operation on the sparse decomposition coefficients corresponding to the atom dictionaries obtained after sparse decomposition and the obtained fault weight matrix, and taking the calculation result as a fault characteristic result.
The present invention is described in detail below with reference to specific data:
a comparison experiment is carried out on the improved sparse decomposition method and the wavelet decomposition method, and the improved sparse decomposition method is proved to be superior to the wavelet decomposition method in extraction of the weak fault characteristics of the AUV propeller. The wavelet decomposition method is also used for extracting the fault characteristics of the AUV propeller.
AUV pool experiment fault sample data used in the patent comparison experiment are 10% output loss fault and 5% output loss fault. The time when the failure occurred in the sample was 200 beats. The following patent takes a 5% loss-of-output fault sample as an example. The sampling beat interval of the signals in the samples is 0.3s, and the sampling frequency is 5 Hz. Since the first 50 beats in the data are AUV starting stage, the data are operated at the steady speed of 0.3m/s from the 51 th beat. And analyzing the experimental data with the beat of 51-550.
The comparison experiment is carried out by taking the speed signal, the heading angle and the main thrust voltage as analysis objects, and the comparison experiment result of the patent and the wavelet decomposition method is shown in figures 3-5. In order to better analyze the lifting effect of the patent, specific data in fig. 3-5 are extracted and the effects of the two methods are compared. The specific data are shown in fig. 6.
The results of the experiments in FIGS. 3-5 and 6 were summarized. In the feature extraction stage, the higher the fault feature value is after feature extraction, the better the effect of the feature extraction method is. Therefore, the patent verifies that the patent is superior to the wavelet decomposition method through the ratio of the fault characteristic to the maximum value of the interference characteristic. By generalizing the results, it can be found that: after the three signals are separated by a wavelet decomposition method, the ratio of the characteristic values is 1.278, 1.138 and 1.676 respectively, and the ratio of the characteristic values of the patent is 1.588, 1.708 and 2.381. It can be found that the ratio of the feature values of the decomposition results of the method and the wavelet decomposition method is greater than 1, which indicates that the feature values at the fault moment are higher than the interference feature values at other moments, i.e. the two methods have certain effect on weak fault feature extraction. Comparing the two methods, the signals of different sensors are respectively improved by 24.25%, 50.08% and 42.06% compared with the wavelet decomposition method. Based on the analysis, the method is superior to the traditional wavelet decomposition method in the aspect of extraction of weak fault characteristics of the thruster.
In summary, the invention relates to a weak fault feature extraction method of an autonomous underwater robot propeller based on combination of sparse decomposition and a fault weight matrix. Belongs to the technical field of underwater robot fault diagnosis. The method firstly adds a time variable to the traditional sparse decomposition algorithm. Each beat of the signal to be decomposed can obtain a corresponding decomposition result, so that the precision of the decomposition result in a time domain is higher; then enhancing fault characteristics on the basis of a traditional characteristic extraction method, and storing fault information in a fault weight matrix by establishing the fault weight matrix corresponding to the atom dictionary; and finally, taking the inner product of the sparse decomposition coefficient and the fault weight matrix as the enhanced fault characteristic, and further increasing the difference value between the fault characteristic and the interference characteristic. The method can effectively highlight and extract fault characteristics, is convenient for subsequent detection and identification of weak faults of the propeller, and is particularly suitable for monitoring the state of the autonomous underwater robot propeller.

Claims (3)

1. A weak fault feature extraction method for a propeller of an autonomous underwater robot is characterized by comprising the following steps: the method comprises the following steps:
step 1: performing atomic dictionary learning, namely performing dictionary learning by using an existing data sample to obtain a basis function similar to the structural characteristics of the AUV state signal;
step 2: sparse decomposition, which is to perform sparse decomposition on the signal based on the learnt atom dictionary and the time variable;
and step 3: continuously carrying out sparse decomposition on the fault sample, and updating a fault weight matrix;
and 4, step 4: and performing inner product operation on the sparse decomposition coefficients corresponding to the atom dictionaries obtained after sparse decomposition and the obtained fault weight matrix, and taking the calculation result as a fault characteristic result.
2. The weak fault feature extraction method of the autonomous underwater robot propeller according to claim 1, characterized in that: the step 2 specifically comprises the following steps:
the set of the atom dictionary at any time tau is obtained by adding a time shift operator, and the expression of the k-th iteration sparse decomposition is as follows:
Figure FDA0002339965100000011
wherein:
Figure FDA0002339965100000012
representing the input signal, TτRepresents the time shift operator, mkRepresenting a collection of atom dictionaries, TτmkI.e. represented as a collection of atom dictionaries at each time instant other than the time instant t, αk,τRepresenting sparse decomposition coefficients, epsilon representing sparse decomposition residual terms；
The corresponding constraints are:
Figure FDA0002339965100000013
in the formula: l0The norm represents the number of non-zero terms in the coefficient vector, is a non-deterministic polynomial, namely cannot be directly solved, and is generally selected from l in practical application2Norm is approximated, so the above equation is further represented by l2The norm is expressed as:
Figure FDA0002339965100000014
the equation for obtaining the best-matched atom and the corresponding sparse decomposition coefficient obtained by the kth iteration in the decomposition process based on the formula is as follows:
Figure FDA0002339965100000015
the above formula represents the most matched atom and the constraint condition satisfied by the corresponding sparse decomposition coefficient in the k-th iteration, and the most matched atom m is finally obtained by solving the above formulakAnd corresponding sparse decomposition coefficients αk,τThe expression of (a) is:
Figure FDA0002339965100000016
Figure FDA0002339965100000021
wherein,
Figure FDA0002339965100000022
representing the time-shift operator Tταk,τRepresents the best matching atom m in the final decomposition resultkCorresponding sparse decomposition coefficients.
3. The weak fault feature extraction method of the autonomous underwater robot propeller as claimed in claim 2, wherein: the step 3 specifically comprises:
1) initialization
Taking J samples to be learned { x1,x2…xJJ is 1, and the residual error term is initialized to be epsilon and xjThe atom dictionary D is an atom dictionary which is learned based on a K-SVD algorithm, the iteration time t is 1, and a coefficient matrix is decomposed
Figure FDA0002339965100000023
Fault weight matrix
Figure FDA0002339965100000024
2) Finding matching atoms
Calculating residual error epsilon and each atom D in atom dictionary DjAnd find the lower subscript λ of the maximum value of the inner product, i.e.:
|<x,dj>|＝supi∈(1,...,k)|<x,dλ>|
3) updating coefficient matrix and fault weight matrix
The decomposition coefficient matrix is updated as:
αt＝αt-1∪|<x,dλ>|
and updating the corresponding fault weight matrix as follows:
Wλ＝Wλ+αt
4) updating residual errors
The residual is the difference between the original signal and the current iteration fitting result, that is:
εt＝x-Dtαt
5) determining termination condition
The iteration termination condition of the algorithm is the number i of atoms used for decomposing the signal, and if the iteration times t is less than i, the step 1 is returned; and if the iteration time t is more than or equal to i, learning the next group of data, namely j is j +1, carrying out normalization processing on the fault weight matrix when all data are completely learned, and then stopping the algorithm.
CN201911372169.9A 2019-12-27 2019-12-27 Weak fault feature extraction method for autonomous underwater robot propeller Active CN111091159B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911372169.9A CN111091159B (en) 2019-12-27 2019-12-27 Weak fault feature extraction method for autonomous underwater robot propeller

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911372169.9A CN111091159B (en) 2019-12-27 2019-12-27 Weak fault feature extraction method for autonomous underwater robot propeller

Publications (2)

Publication Number Publication Date
CN111091159A true CN111091159A (en) 2020-05-01
CN111091159B CN111091159B (en) 2023-03-17

Family

ID=70397400

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911372169.9A Active CN111091159B (en) 2019-12-27 2019-12-27 Weak fault feature extraction method for autonomous underwater robot propeller

Country Status (1)

Country Link
CN (1) CN111091159B (en)

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080239874A1 (en) * 2005-08-03 2008-10-02 Kemp Mathieu P Underwater guidance systems, unmanned underwater vehicles and methods
US20110297070A1 (en) * 2008-11-04 2011-12-08 Riggs Neil P Propulsion System for an Autonomous Underwater Vehicle
CN104503432A (en) * 2014-11-27 2015-04-08 哈尔滨工程大学 Autonomous underwater robot fault identification method based on wavelet energy
US20150213323A1 (en) * 2014-01-27 2015-07-30 Xerox Corporation Video anomaly detection based upon a sparsity model
CN107132760A (en) * 2017-03-27 2017-09-05 哈尔滨工程大学 The underwater robot state monitoring method described based on fuzzy support vector domain
CN107992843A (en) * 2017-12-13 2018-05-04 武汉科技大学 A kind of sparse features extracting method of rotating machinery fault signal
US20180231394A1 (en) * 2014-09-10 2018-08-16 Siemens Energy, Inc. Gas turbine sensor failure detection utilizing a sparse coding methodology
CN109033965A (en) * 2018-06-22 2018-12-18 江苏科技大学 A kind of underwater robot propeller failure time-frequency characteristics Enhancement Method
CN109634308A (en) * 2019-01-16 2019-04-16 中国海洋大学 Based on intelligent navigation method under dynamic (dynamical) rate pattern auxiliary water
CN109993105A (en) * 2019-03-29 2019-07-09 北京化工大学 A kind of improved adaptive sparse sampling Fault Classification

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080239874A1 (en) * 2005-08-03 2008-10-02 Kemp Mathieu P Underwater guidance systems, unmanned underwater vehicles and methods
US20110297070A1 (en) * 2008-11-04 2011-12-08 Riggs Neil P Propulsion System for an Autonomous Underwater Vehicle
US20150213323A1 (en) * 2014-01-27 2015-07-30 Xerox Corporation Video anomaly detection based upon a sparsity model
US20180231394A1 (en) * 2014-09-10 2018-08-16 Siemens Energy, Inc. Gas turbine sensor failure detection utilizing a sparse coding methodology
CN104503432A (en) * 2014-11-27 2015-04-08 哈尔滨工程大学 Autonomous underwater robot fault identification method based on wavelet energy
CN107132760A (en) * 2017-03-27 2017-09-05 哈尔滨工程大学 The underwater robot state monitoring method described based on fuzzy support vector domain
CN107992843A (en) * 2017-12-13 2018-05-04 武汉科技大学 A kind of sparse features extracting method of rotating machinery fault signal
CN109033965A (en) * 2018-06-22 2018-12-18 江苏科技大学 A kind of underwater robot propeller failure time-frequency characteristics Enhancement Method
CN109634308A (en) * 2019-01-16 2019-04-16 中国海洋大学 Based on intelligent navigation method under dynamic (dynamical) rate pattern auxiliary water
CN109993105A (en) * 2019-03-29 2019-07-09 北京化工大学 A kind of improved adaptive sparse sampling Fault Classification

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
TULV等: "Fault feature extraction method based on optimized sparse decomposition algorithm for AUV with weak thruster fault", 《OCEAN ENGINEERING》 *
ZHANG, M.等: "Thruster fault feature extraction for autonomous underwater vehicle in time-varying ocean currents based on single channel blind source separation", 《PROCEEDINGS OF THE INSTITUTION OF MECHANICAL ENGINEERS》 *
姜春萌: "自主水下机器人智能控制与参数优化技术研究", 《中国优秀博硕士学位论文全文数据库(博士)工程科技Ⅱ辑》 *
张菀: "基于信号稀疏表示的滚动轴承微弱故障特征提取方法研究", 《中国优秀博硕士学位论文全文数据库(博士)工程科技Ⅱ辑》 *
张铭钧等: "自主式水下机器人故障特征增强方法", 《哈尔滨工程大学学报》 *
朱晨光: "基于优化随机共振系统和灰色关联理论的AUV推进器弱故障诊断方法研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 *

Also Published As

Publication number Publication date
CN111091159B (en) 2023-03-17

Similar Documents

Publication Publication Date Title
Fišerová et al. On the interpretation of orthonormal coordinates for compositional data
CN109033965B (en) Underwater robot propeller fault time-frequency feature enhancement method
CN113176092B (en) Motor bearing fault diagnosis method based on data fusion and improved experience wavelet transform
CN113255733A (en) Unsupervised anomaly detection method under multi-modal data loss
CN107132760B (en) Underwater robot state monitoring method based on fuzzy support vector field description
CN111665050B (en) Rolling bearing fault diagnosis method based on clustering K-SVD algorithm
CN111476339B (en) Rolling bearing fault feature extraction method, intelligent diagnosis method and system
CN116243682B (en) Underwater propeller fault detection method based on improved generation of countermeasure network
CN115200850A (en) Mechanical equipment anomaly detection method under explicit representation of multi-point sample structure information
CN113642508A (en) Bearing fault diagnosis method based on parameter self-adaptive VMD and optimized SVM
CN112560699A (en) Gear vibration information source underdetermined blind source separation method based on density and compressed sensing
CN113468760A (en) Motor weak fault detection method and system based on dictionary learning
CN111091159B (en) Weak fault feature extraction method for autonomous underwater robot propeller
CN116361723A (en) Bearing fault diagnosis and classification method based on multi-scale characteristics and attention
CN114943257A (en) Synchronous noise reduction and disturbance classification identification method of distributed optical fiber sensing system
CN113639985A (en) Mechanical fault diagnosis and state monitoring method based on optimized fault characteristic frequency spectrum
CN117388711A (en) Combined estimation method, device, equipment and medium for battery electric quantity and health degree
CN111144307A (en) Fault feature extraction method for autonomous underwater robot propeller
CN116340827A (en) Underwater propeller fault diagnosis method and system based on predicted beat dynamic adjustment
CN116432081A (en) Migration diagnosis method based on integrated weighted subdomain self-adaptive network
CN115753107A (en) Weak rolling bearing fault diagnosis method based on nonlinear quantum information entropy
CN114706128A (en) Microseism data noise suppression method and system, storage medium and seismic information processing equipment
CN116028858A (en) Pre-labeled self-supervision neural network learning heart beat classification method and system
CN114565850A (en) Hyperspectral image abnormal target detection method based on total variation low-rank tensor decomposition
CN114200232A (en) Method and system for detecting fault traveling wave head of power transmission line

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