WO2000070310A1 - Dispositif d'identification de signal faisant intervenir un algorithme genetique et systeme d'identification en ligne - Google Patents

Dispositif d'identification de signal faisant intervenir un algorithme genetique et systeme d'identification en ligne Download PDF

Info

Publication number
WO2000070310A1
WO2000070310A1 PCT/JP2000/003006 JP0003006W WO0070310A1 WO 2000070310 A1 WO2000070310 A1 WO 2000070310A1 JP 0003006 W JP0003006 W JP 0003006W WO 0070310 A1 WO0070310 A1 WO 0070310A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal
feature parameters
identification
feature
state
Prior art date
Application number
PCT/JP2000/003006
Other languages
English (en)
Japanese (ja)
Inventor
Ho Jinyama
Toshio Toyota
Takashi Nishimura
Original Assignee
Kyushu Kyohan 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 Kyushu Kyohan Co., Ltd. filed Critical Kyushu Kyohan Co., Ltd.
Priority to GB0100769A priority Critical patent/GB2358253B8/en
Publication of WO2000070310A1 publication Critical patent/WO2000070310A1/fr

Links

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/02—Feature extraction for speech recognition; Selection of recognition unit
    • 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
    • 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
    • G06F18/211—Selection of the most significant subset of 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

Definitions

  • the present invention relates to an identification device and an online identification system for identifying a feature or state of a signal in equipment diagnosis, voice recognition, medical diagnosis or pattern recognition, and the like.
  • the identification accuracy is measured Depending on the range of the device, the type of sensor, and the environment of the target object, it is difficult to identify the characteristics of the signal shape.
  • linear discriminant function method is effective only when linear classification can be performed between groups of patterns (states) to be identified.
  • form of the function In the case of a nonlinear discriminant function, the form of the function must be determined in advance, but it is not known in advance which form is better. Furthermore, in the case of using a neural network for identification, if the pattern becomes complicated, there is no effective method for determining the structure of the neural network or guaranteeing learning convergence. Disclosure of the invention
  • the characteristics of the signal are determined by using a genetic algorithm (GA) (see the above-mentioned reference (3)). Automatically generate good GA feature parameters for identification.
  • GA genetic algorithm
  • GA feature parameters Self-organization of primitive feature parameters is performed automatically using a genetic algorithm (GA), and new feature parameters (hereinafter referred to as GA feature parameters) are automatically generated.
  • GA genetic algorithm
  • the calculated GA feature parameters are evaluated using the “identification index DI” or “similarity index SI”, and if they are good feature parameters, they are used for identification. Start over from (2).
  • the probability distribution function and the judgment criterion for identification are determined by probability theory, Dempster & Shafer probability theory, and probability theory.
  • the characteristics of the signal are identified online using the GA characteristic parameter for the signal measured by the permanently installed sensor.
  • step (3) good feature parameters can be quickly found in step (3) above, and it is almost never necessary to repeatedly define primitive feature parameters. Since it is performed in an efficient manner, the extraction work of GA feature parameters for identification is efficient.
  • G Genetic algorithm
  • GP genetic programming
  • FIG. 1 is a flowchart showing the flow of automatic generation of GA feature parameters.
  • Figure 2 is a flow chart showing the flow of signal characteristics or state identification by the GA diagnostic device.
  • Figure 3 is a flow chart showing the flow of identification of signal features or states online.
  • Fig. 4 is a flowchart showing the flow of sequential generation of the GA feature parameters and signal feature identification.
  • FIG. 5 is an explanatory diagram showing an example of a tree graph.
  • Figure 6 shows the discrimination rate P for state 2.
  • 6 is a graph showing an example of the above.
  • FIG. 7 is an explanatory diagram showing crossover and mutation.
  • FIG. 8 is a graph showing an example of a vibration signal between a normal state and a misalignment.
  • Figure 9 is a table showing the relationship between the discrimination index of the primitive feature parameters and the discrimination rate.
  • Figure 10 is a table showing the relationship between the discrimination index and the discrimination rate for the GA feature parameters.
  • FIG. 11 is a graph showing a vibration signal and a spectrum in each state of the gear.
  • Figure 12 is a table showing the relationship between the identification index of the atomic feature parameters and the identification rate.
  • FIG. 13 is a graph showing an example of the probability distribution function.
  • FIG. 14 is a front view showing an example of a panel of the GA discriminator.
  • Figure 15 is a graph showing examples of the probability distribution functions for normal, cautionary, and dangerous states.
  • FIG. 16 is a graph showing a probability distribution function approximated by a straight line.
  • Figure 17 is an example of a “trend management graph” of states according to the degree of possibility.
  • FIG. 18 is a circuit diagram showing an example of a circuit of the GA discriminator. BEST MODE FOR CARRYING OUT THE INVENTION
  • 1 is the discrete value of (t) after the 80 transformation
  • ⁇ is the mean and standard deviation of X i, respectively.
  • x P is the average value of the maximum value of the waveform (peak value), / x (5)
  • primitive feature parameters can be defined in addition to the above primitive feature parameters, but when applying this method, first try the above primitive feature parameters, and if the effect is not good, Further, other primitive feature parameters may be arbitrarily defined and executed again.
  • Isseki Using GA, to automatically generate a GA feature parameters Isseki p CA by reorganization of m primitives parameters Isseki proposes the following various methods.
  • the difference between pGA and the conventional linear discriminant function is that C i is determined by GA.
  • f ( ⁇ ) represents an arbitrary mathematical expression, which is determined by GA.
  • Equations (15) and (17) are linear and nonlinear combinations of primitive feature parameters, respectively. You. Since the coefficient C i is considered to be a weight for each primitive feature parameter, if the values of each primitive feature parameter obtained from the target signal are normalized as follows, then equation (16) The GA feature parameter shown in is equivalent to equation (15) or equation (17).
  • FIG. 5 is an example of a tree graph.
  • the operators of the four arithmetic operations (10, 1, *, /, n) are put in the “sections” of the tree graph, and the primitive feature parameters ( ⁇ ⁇ , P2, ⁇ ) are put in the “leaves”. .
  • the fitness of the gene for identifying the two-state signal that is, the goodness of the feature parameter, is evaluated by the following formula. (1 8)
  • DI Discrimination Index
  • ⁇ And 2 are the average values of the feature parameters found in states 1 and 2, and ⁇ i and ⁇ 2 are their standard deviations.
  • the discrimination rate (Distinction Rate, D R) for the feature parameters for state 1 and state 2 can be obtained by the following equation. (i 9)
  • DI can be used as an index (fitness) for evaluating the goodness of the feature parameters.
  • the population of early genes is randomly generated.
  • the generation algorithm is as follows.
  • Crossover of genes uses single-point crossover or multipoint crossover.
  • two genes are selected from the gene population of the current generation according to the probability ⁇ in the following equation and crossed.
  • Figure 7 is an example of one-point intersection of two tree graphs. In this case, for each tree, the part below the “node” (1) arbitrarily selected is exchanged (crossed).
  • Mutations are made at the rate of mutation at the “nodes” (10, 1, *, /, n) or “leaves” (pi) of the tree.
  • FIG. 7 shows examples of crossovers and mutations.
  • the fitness (D I) is greater than a preset value. For example, when D I> 3, the identification rate is P. Is greater than 99. 99%.
  • FIG. 9 shows the discrimination index DI and discrimination rate DR. It can be seen that the discriminating ability of these primitive feature parameters is not high.
  • Figure 10 shows some automatically generated GA feature parameters. It can be seen that the DI and DR of these GA feature parameters were greatly improved over the DI and DR of primitive feature parameters.
  • Figure 11 shows an example of the signal.
  • g (t) is an acceleration time-series signal of each state
  • F (f) is a spectrum of each state
  • S (f) is a normal spectrum ratio.
  • the meaning of each symbol is as follows.
  • gw (t), Fw (f), Sw (f) Fully worn state of the teeth (hereafter, worn state)
  • the waveform of the signal changes depending on other abnormal states and the degree of the abnormal state, but an attempt is made to diagnose.
  • the proposed method can extract good GA feature parameters that can similarly identify the characteristics of the signal.
  • a normal signal can be considered noise. That is, Eliminating the normal signal from the signal measured at the time of diagnosis will result in an abnormal signal, and its characteristics will be relatively easy to identify.
  • the frequency domain of the abnormal signal is unknown, and even if it is known, there is no generally effective noise elimination method if the correlation with noise is strong. The procedure is as follows.
  • Step 2 In order to remove the effect of the spectrum gain, find the normalized spectrum as follows.
  • the characteristic frequency components clearly appear in the normal spectrum ratios S e (f), S m (f), and S w (f) obtained in each abnormal state. Understand. This is the same effect as noise removal.
  • Table 1 shows the discrimination index DI and the discrimination rate DR for each characteristic parameter when the gear failure is identified by the seven characteristic parameters shown in Equations (25) to (31).
  • Figures 12 (a), (b), and (c) show the discrimination index DI and discrimination rate for discriminating between the normal state and the eccentric state, the normal state and the misalignment state, and the normal state and the wear state for p! To p, respectively.
  • the maximum discrimination rate for each is 9 1% (p 5 ), 95.6% (p 2 ), and 93.1% (ps), indicating that these parameters are not enough to identify each fault condition.
  • Equations (25) to (31) there are various characteristic parameters, but it is not easy to easily find good characteristic parameters. Therefore, GA is used to automatically generate GA features with high identification accuracy by self-reorganization of arbitrarily defined primitive features such as Eqs. (25) to (31). Generated.
  • the main settings for searching for the GA characteristic parameters that distinguish between the normal state of the gear and each abnormal state are as follows.
  • the number of genes in one generation is 200.
  • the maximum number of layers in the tree should be 6 or less. If it exceeds 6, start over.
  • Equations (25) to (31) are the normal and eccentric states, the normal and misalignment states, and The search results (optimal GA feature parameters) after 100 generations of evolution are shown to distinguish between normal and wear conditions.
  • the identification parameters DI, 6.01 in the eccentric state, 5.17 in the misalignment state, 9.04 in the worn state
  • the identification parameters for identifying each state in the feature parameters are shown in Fig. 12. It is much better than the indicator, and its identification rate has reached almost 100%. If the above settings and the maximum number of evolutionary generations are changed, the finally obtained GA feature parameter formula will be different, but if the identification rate is sufficiently high, it can be used for abnormality diagnosis.
  • the signal in that state is divided into two groups ( X l (t) and ⁇ 2 (t)), and the similarity index (Similarity index) defined as Index, SI) ”is used.
  • SI similarity index
  • the SI calculation formula can be expressed as follows.
  • SI f ⁇ , 2 , ox, ⁇ 2 , ⁇ ⁇ 2, ⁇ i, ⁇ 2 ) (36)
  • the subscripts 1 and 2 indicate the values obtained from (t) and ⁇ 2 (t), respectively. .
  • D is defined as follows.
  • Wi is a weighting factor
  • C and C are constants.
  • SI 2 ( ⁇ 1 + ⁇ 2 ) / D (3 9)
  • SI 4 1 / (W, ⁇ I + W 2 G 2 + W3D) (4 0)
  • W>, W 2 and W 3 are weighting factors.
  • the probability distribution function ⁇ (X) is obtained from the probability density function f k (x) of X using the measured data in state k by Eq. (41).
  • the possibility theory (4) L. Davis: HANDBOOK OF GENETIC ALGORITHMS, Van Nostrand Reinhold, A Division of Wadsworth, Inc (1990)
  • it is possible for x to follow any probability distribution A gender distribution function is determined, and a membership function for identification can be created based on the likelihood distribution function. For example, if X follows a normal distribution, the N-stage probability distribution function p k ( X i ) is obtained as follows.
  • the state k the possibilities w y is determined as follows.
  • Figure 14 shows an example of the panel display of the GA classifier.
  • the state is determined by obtaining the probability distribution function p n (x) of the GA feature parameter X for identification, then the left and right sides
  • the probability distribution functions (p dl (x) and p d2 (x)) of the “dangerous state” are determined as shown in Figure 15.
  • the possibility of “normal”, “caution”, and “danger” obtained during actual identification is displayed on the panel of the GA classifier as shown in Fig. 14 (a).
  • each possibility distribution function can be approximated by a triangle or a trapezoid as shown in FIG.
  • the tendency management of the state according to the degree of possibility is performed by creating a “trend management graph” as shown in Fig. 17 (for equipment diagnosis).
  • Fig. 18 shows an example of a circuit diagram of a GA classifier.
  • 1 is a sensor
  • 2 is a charge amplifier
  • 3 is a filter module
  • 4 is a one-chip CPU
  • 5 is a result display
  • 6 is data RAM
  • 7 is an AD converter
  • 8 is a DC port
  • 9 is SCI.
  • 10 is 0 PU and 1 is flash ROM.
  • the accuracy of abnormality identification is improved by a classifier using good GA feature parameters searched by GA and a probability distribution function for identification.
  • the computer and the GA classifier automatically perform the process from the creation of the feature parameters to the display of the status, so that even non-experts without specialized knowledge can easily perform the classification.
  • the present invention can be used in fields such as equipment diagnosis, voice recognition, medical diagnosis, and pattern recognition.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Biophysics (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Computation (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Human Computer Interaction (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Physiology (AREA)
  • Genetics & Genomics (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

Cette invention concerne la définition d'un paramètres de traits (paramètre de traits primitifs),l'auto-réorganisation de ce paramètre au moyen d'un algorithme génétique (GA) et la génération d'un nouveau paramètre de traits (paramètre de traits GA). Si le résultat d'une évaluation du paramètre de traits GA au moyen d'un indice de distinction ou d'un indice d'évaluation est acceptable, on applique une théorie des probabilités, la théorie des probabilités de Dempster & Shafer (DS) et une fonction de possibilité au paramètre de traits GA pour générer une fonction de distribution de possibilité et un critère de jugement à des fins de distinction.
PCT/JP2000/003006 1999-05-12 2000-05-10 Dispositif d'identification de signal faisant intervenir un algorithme genetique et systeme d'identification en ligne WO2000070310A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
GB0100769A GB2358253B8 (en) 1999-05-12 2000-05-10 Signal identification device using genetic algorithm and on-line identification system

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP13193799 1999-05-12
JP11/131937 1999-05-12

Publications (1)

Publication Number Publication Date
WO2000070310A1 true WO2000070310A1 (fr) 2000-11-23

Family

ID=15069706

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2000/003006 WO2000070310A1 (fr) 1999-05-12 2000-05-10 Dispositif d'identification de signal faisant intervenir un algorithme genetique et systeme d'identification en ligne

Country Status (2)

Country Link
GB (1) GB2358253B8 (fr)
WO (1) WO2000070310A1 (fr)

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002372440A (ja) * 2001-06-14 2002-12-26 Ho Jinyama 状態判定法並びに状態判定装置及び状態判定機能を備えた信号収録装置
JP2003317083A (ja) * 2002-04-25 2003-11-07 Dainippon Screen Mfg Co Ltd 画像分類方法、プログラムおよび画像分類装置
WO2004068078A1 (fr) * 2003-01-10 2004-08-12 Ho Jinyama Procede d'estimation d'etat, procede et dispositif de prediction d'etat
JP2004279211A (ja) * 2003-03-14 2004-10-07 Omron Corp 知識作成支援装置及びパラメータ探索方法並びにプログラム製品
JP2004309998A (ja) * 2003-02-18 2004-11-04 Nec Corp 確率分布推定装置および異常行動検出装置，ならびにその確率分布推定方法および異常行動検出方法
JP2006072659A (ja) * 2004-09-01 2006-03-16 Matsushita Electric Works Ltd 信号識別方法および信号識別装置
JP2007051982A (ja) * 2005-08-19 2007-03-01 Japan Science & Technology Agency 診断対象物の評価方法および評価装置
JP2009503533A (ja) * 2005-08-05 2009-01-29 コーニンクレッカ フィリップス エレクトロニクス エヌ ヴィ 動的遺伝子分布によるサーチ空間保護
US7809528B2 (en) 2007-09-11 2010-10-05 Shenzhen Mindray Bio-Medical Electronics Co., Ltd. Waveform analyzing method and apparatus for physiological parameters
CN106228562A (zh) * 2016-08-01 2016-12-14 浙江科技学院 基于概率神经网络算法的在线印刷品色彩质量评价方法
JP2017194371A (ja) * 2016-04-21 2017-10-26 株式会社トクヤマ 回転駆動装置における診断対象部の異常診断方法と、それに用いる異常診断装置

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101165779B (zh) * 2006-10-20 2010-06-02 索尼株式会社 信息处理装置和方法、程序及记录介质
JP4239109B2 (ja) 2006-10-20 2009-03-18 ソニー株式会社 情報処理装置および方法、プログラム、並びに記録媒体
US20150179167A1 (en) * 2013-12-19 2015-06-25 Kirill Chekhter Phoneme signature candidates for speech recognition

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH0993665A (ja) * 1995-09-26 1997-04-04 Meidensha Corp 監視装置

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH0993665A (ja) * 1995-09-26 1997-04-04 Meidensha Corp 監視装置

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
CHIN et al., "Method of Failure Detection by Possibility Theory and Dempster & Shafer Probability Theory", Journal of Kita Kyushu Iko Gakujutsu Kyokai, Vol. 9, No. 1, 1998, pages 1-4. *
CHIN et al., "Nuclear Reactor Diagnostic System Using GA Trained Neural Networks", Transactions C of IEEJ, Vol. 114-C, No. 9, August 1994, pages 922-929. *
SO et al., "Diagnosis Method for a Gear Equipment by Sequential Fuzzy Neural Network", Journal of Nippon Setsubi Kanri Gakkai, Vol. 10, No. 1, 1998, pages 15-20. *
Y. ICHIKAWA, "Identeki Algorithm no Seigyokei Sekkei eno Ouyou", Transactions D of IEEJ, Vol. 113-D, No. 12, December 1993, pages 1361-1362. *

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002372440A (ja) * 2001-06-14 2002-12-26 Ho Jinyama 状態判定法並びに状態判定装置及び状態判定機能を備えた信号収録装置
JP2003317083A (ja) * 2002-04-25 2003-11-07 Dainippon Screen Mfg Co Ltd 画像分類方法、プログラムおよび画像分類装置
WO2004068078A1 (fr) * 2003-01-10 2004-08-12 Ho Jinyama Procede d'estimation d'etat, procede et dispositif de prediction d'etat
US7561991B2 (en) 2003-02-18 2009-07-14 Nec Corporation Detection of abnormal behavior using probabilistic distribution estimation
JP2004309998A (ja) * 2003-02-18 2004-11-04 Nec Corp 確率分布推定装置および異常行動検出装置，ならびにその確率分布推定方法および異常行動検出方法
JP2004279211A (ja) * 2003-03-14 2004-10-07 Omron Corp 知識作成支援装置及びパラメータ探索方法並びにプログラム製品
JP2006072659A (ja) * 2004-09-01 2006-03-16 Matsushita Electric Works Ltd 信号識別方法および信号識別装置
JP2009503533A (ja) * 2005-08-05 2009-01-29 コーニンクレッカ フィリップス エレクトロニクス エヌ ヴィ 動的遺伝子分布によるサーチ空間保護
JP2007051982A (ja) * 2005-08-19 2007-03-01 Japan Science & Technology Agency 診断対象物の評価方法および評価装置
US7809528B2 (en) 2007-09-11 2010-10-05 Shenzhen Mindray Bio-Medical Electronics Co., Ltd. Waveform analyzing method and apparatus for physiological parameters
US8000937B2 (en) 2007-09-11 2011-08-16 Shenzhen Mindray Bio-Medical Electronics Co., Ltd. Method and apparatus for waveform analysis of physiological parameters
JP2017194371A (ja) * 2016-04-21 2017-10-26 株式会社トクヤマ 回転駆動装置における診断対象部の異常診断方法と、それに用いる異常診断装置
CN106228562A (zh) * 2016-08-01 2016-12-14 浙江科技学院 基于概率神经网络算法的在线印刷品色彩质量评价方法

Also Published As

Publication number Publication date
GB2358253B (en) 2003-09-03
GB0100769D0 (en) 2001-02-21
GB2358253A (en) 2001-07-18
GB2358253B8 (en) 2011-08-03

Similar Documents

Publication Publication Date Title
WO2000070310A1 (fr) Dispositif d'identification de signal faisant intervenir un algorithme genetique et systeme d'identification en ligne
Yang et al. Fault diagnosis of induction motor based on decision trees and adaptive neuro-fuzzy inference
CN106338406B (zh) 列车牵引电传动系统的在线监测与故障预警系统及方法
CN110647943B (zh) 基于演化数据聚类分析的切削刀具磨损监测方法
Wu et al. Induction machine fault detection using SOM-based RBF neural networks
JP2008059601A (ja) 双峰性データを判別する方法
Juniati et al. Fractal dimension to classify the heart sound recordings with KNN and fuzzy c-mean clustering methods
Azami et al. An improved signal segmentation method using genetic algorithm
CN102788696A (zh) 基于改进bp神经网络和模糊集理论的轴承健康度评价方法
Al Tobi et al. Faults diagnosis of a centrifugal pump using multilayer perceptron genetic algorithm back propagation and support vector machine with discrete wavelet transform‐based feature extraction
CN113095151B (zh) 基于信号分解与复杂网络的滚动轴承未知故障检测方法
CN114896228B (zh) 基于过滤规则多级组合优化的工业数据流清洗模型和方法
CN105306438B (zh) 基于模糊粗糙集的网络安全态势评估方法
Al Tobi et al. Using MLP‐GABP and SVM with wavelet packet transform‐based feature extraction for fault diagnosis of a centrifugal pump
CN118296470B (zh) 一种多分类器融合的智能分类方法
EP3581098B1 (fr) Procédé de génération de classificateur d'état respiratoire et procédé de détermination d'état respiratoire
CN115471043A (zh) 一种基于集成sae-som的蜡模铸造热处理设备健康状态评估方法
CN117808600A (zh) 证券交易数据分析方法及装置
Liao et al. A vibration signal filtering method based on KL divergence genetic algorithm–with application to low speed bearing fault diagnosis
Yildirim et al. Heartbeat type classification with optimized feature vectors
CN115310499B (zh) 一种基于数据融合的工业设备故障诊断系统及方法
CN116089897A (zh) 一种基于多源信息融合的电力设备运行状态评估方法及系统
Laosiritaworn et al. Classification techniques for control chart pattern recognition: A case of metal frame for actuator production
JPH07219581A (ja) 音響信号の識別方法および装置
Osowski et al. Analysis of features for efficient ECG signal classification using neuro-fuzzy network

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): GB JP US

ENP Entry into the national phase

Ref country code: JP

Ref document number: 2000 618695

Kind code of ref document: A

Format of ref document f/p: F

ENP Entry into the national phase

Ref country code: GB

Ref document number: 200100769

Kind code of ref document: A

Format of ref document f/p: F

WWE Wipo information: entry into national phase

Ref document number: 09743362

Country of ref document: US