CN108573278A - A kind of multiple operating modes process modal identification method - Google Patents

A kind of multiple operating modes process modal identification method Download PDF

Info

Publication number
CN108573278A
CN108573278A CN201810207588.6A CN201810207588A CN108573278A CN 108573278 A CN108573278 A CN 108573278A CN 201810207588 A CN201810207588 A CN 201810207588A CN 108573278 A CN108573278 A CN 108573278A
Authority
CN
China
Prior art keywords
data
operating mode
probability density
operating modes
under
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
CN201810207588.6A
Other languages
Chinese (zh)
Other versions
CN108573278B (en
Inventor
周东华
尚骏
陈茂银
张海峰
李明亮
卢晓
钟麦英
王建东
王友清
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.)
Shandong University of Science and Technology
Original Assignee
Shandong University of Science and Technology
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 Shandong University of Science and Technology filed Critical Shandong University of Science and Technology
Priority to CN201810207588.6A priority Critical patent/CN108573278B/en
Publication of CN108573278A publication Critical patent/CN108573278A/en
Application granted granted Critical
Publication of CN108573278B publication Critical patent/CN108573278B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)
  • Complex Calculations (AREA)

Abstract

The invention discloses a kind of multiple operating modes process modal identification methods, belong to automatic measurement technique field, including off-line training and online modal identification two parts；Off-line training calculates the weight of each mode using historic training data, and it is obtained by orthogonal transformation so that the conditional sampling degree of variable is maximum under different operating modes, its orthogonal matrix is obtained by numerical method solution optimization problem, Bayes classifier is trained using the off-line data after transformation, utilizes kernel function estimation conditional probability density function；Online modal identification carries out orthogonal transformation to sample first, and calculate the conditional probability density function under each operating mode, mode classification is determined by maximization conditional probability density function, compared with classical way naive Bayesian, it is independent it is assumed that having higher classification accuracy that institute's extracting method of the present invention relaxes data qualification.

Description

A kind of multiple operating modes process modal identification method
Technical field
The invention belongs to automatic measurement technique fields, and in particular to a kind of multiple operating modes process modal identification method.
Background technology
In actual industrial system, operating mode switching can occur because of several factors for the operating mode of production process.These factors include The change of system feeding, the change of reactant ingredient, different production technologies, the variation of external environmental factor and production target Change etc..In some cases, the change of process operating mode has label, as production target change or factory in Operative employee changes charging or change set temperature etc. according to instruction.But the change of some operating modes is not no label, such as external rings The variation of border factor.It is subsequent in this case it is necessary to measurement data analyze the mode of simultaneously on-line identification process Process monitoring provides basis.
Invention content
For the above-mentioned technical problems in the prior art, the present invention proposes a kind of multiple operating modes process modal identification side Method, reasonable design overcome the deficiencies in the prior art, have good effect.
To achieve the goals above, the present invention adopts the following technical scheme that：
A kind of multiple operating modes process modal identification method, specifically comprises the following steps：
Step 1：Off-line training is as follows：
Step 1.1：The historical data under d kind difference operating modes is collected, training dataset is established
Wherein, m is number of probes, niFor the number of samples under i-th of operating mode；
Step 1.2：Calculate the priori weight beta of each operating modei：
Step 1.3：Data under i-th of operating mode are pre-processed：
Wherein,Represent the measurement vector of k-th of sampling instant under i-th of operating mode；For corresponding zero averaging Vector；μ{i}For the sample average of i-th of floor data, calculated by formula (3)：
Step 1.4：Calculate the sample covariance Q of i-th of floor data{i}：
Wherein,ForTransposition；
Step 1.5：Enable j=1, and step-up error threshold value ∈；
Step 1.6：K=0 is enabled, random vector is generatedAnd it normalizes：
Wherein,It indicates2 norms；
Step 1.7：If j>1, then in formula (5)It is orthogonalized processing：
Wherein, wiIt is the i-th row of orthogonal matrix W to be optimized；Middle k is the kth step of iterative calculation, and T is transposition；
Step 1.8：Calculate function g (wj)：
Wherein, wjIt is the jth row of orthogonal matrix W to be optimized；
Step 1.9：Calculate g (wj) Jacobian matrix J (wj)：
Wherein, I is unit battle array；
Step 1.10：Newton iteration：
Step 1.11：If j>1, in formula 9It is orthogonalized processing：
Step 1.12：To in formula 10It is normalized：
Step 1.13：In judgment formula (11), ifEnable k=k+1, and jump to step 1.10 into Row Newton iteration；
Step 1.14：J=j+1 is enabled, if j≤m, jumps to step 1.6；
Step 1.15：Orthogonal matrix W is obtained, the data of different operating modes are handled as follows：
z{i}=WTx{i}(27)；
Wherein, z{i}For the data vector under i-th of operating mode；
Step 1.16：Kernel function K and bandwidth h is set；
Step 1.17：To i=1 ..., d, j=1 ..., m estimate conditional probability density function：
Wherein,Indicate z under i-th of operating mode{i}The value of k-th of sampling instant of j-th of sensor；
Step 2：Online modal identification
Step 2.1：Following orthogonal transformation is carried out to sample data：
zk=WTxk(29)；
Step 2.2：To i=1 ..., d, design conditions probability density
Step 2.3：Mode classification is determined by maximal condition probability density：
Advantageous effects caused by the present invention：
It is compared based on Nae Bayesianmethod with classical, this method improves variable to measurand using orthogonal transformation Conditional sampling degree so that be unsatisfactory for the measurement data of conditional independence assumption originally and to a certain extent more meet to assume. Simultaneously as the shape feature of orthogonal transformation not change data, will not destroy the feature conducive to classification.This method can be used for multiplexing The modal identification problem of condition process can relatively accurately judge unknown operating mode, facilitate subsequent fault detection and diagnosis.
Description of the drawings
Fig. 1 is the flow chart of off-line training according to an embodiment of the invention；
Fig. 2 is the flow chart of online modal identification according to an embodiment of the invention.
Specific implementation mode
Below in conjunction with the accompanying drawings and specific implementation mode invention is further described in detail：
A kind of multiple operating modes process modal identification method, specifically comprises the following steps：
Step 1：Off-line training, flow are as shown in Figure 1；It is as follows：
Step 1.1：The historical data under d kind difference operating modes is collected, training dataset is established
Wherein, m is number of probes, niFor the number of samples under i-th of operating mode；
Step 1.2：Calculate the priori weight beta of each operating modei：
Step 1.3：Data under i-th of operating mode are pre-processed：
Wherein,Represent the measurement vector of k-th of sampling instant under i-th of operating mode；For corresponding zero averaging Vector；μ{i}For the sample average of i-th of floor data, calculated by formula (3)：
Step 1.4：Calculate the sample covariance Q of i-th of floor data{i}：
Wherein,ForTransposition；
Step 1.5：Enable j=1, and step-up error threshold value ∈；
Step 1.6：K=0 is enabled, random vector is generatedAnd it normalizes：
Wherein,It indicates2 norms；
Step 1.7：If j>1, then in formula (5)It is orthogonalized processing：
Wherein, wi is the i-th row of orthogonal matrix W to be optimized；Middle k is the kth step of iterative calculation, and T is transposition；
Step 1.8：Calculate function g (wj)：
Wherein, wjIt is the jth row of orthogonal matrix W to be optimized；
Step 1.9：Calculate g (wj) Jacobian matrix J (wj)：
Wherein, I is unit battle array；
Step 1.10：Newton iteration：
Step 1.11：If j>1, in formula 9It is orthogonalized processing：
Step 1.12：To in formula 10It is normalized：
Step 1.13：In judgment formula (11), ifEnable k=k+1, and jump to step 1.10 into Row Newton iteration；
Step 1.14：J=j+1 is enabled, if j≤m, jumps to step 1.6；
Step 1.15：Orthogonal matrix W is obtained, the data of different operating modes are handled as follows：
z{i}=WTx{i}(42)；
Wherein, z{i}For the data vector under i-th of operating mode；
Step 1.16：Kernel function K and bandwidth h is set；
Step 1.17：To i=1 ..., d, j=1 ..., m estimate conditional probability density function：
Wherein,Indicate z under i-th of operating mode{i}The value of k-th of sampling instant of j-th of sensor；
Step 2：Online modal identification, flow are as shown in Figure 2；It is as follows：
Step 2.1：Following orthogonal transformation is carried out to sample data：
zk=WTxk(44)；
Step 2.2：To i=1 ..., d, design conditions probability density
Step 2.3：Mode classification is determined by maximal condition probability density：
This method and Nae Bayesianmethod are applied to tool there are three in the modal identification of the multiple linear process of operating mode, Compare its classification accuracy.Use 1000 Monte Carlo Experiments.This method and Nae Bayesianmethod classification accuracy it is equal Value is respectively 0.931 and 0.879.Compared with Nae Bayesianmethod, the method for the present invention improves mode in multiple operating modes process and distinguishes The accuracy rate of knowledge demonstrates the validity of proposition method of the present invention.
Certainly, above description is not limitation of the present invention, and the present invention is also not limited to the example above, this technology neck The variations, modifications, additions or substitutions that the technical staff in domain is made in the essential scope of the present invention should also belong to the present invention's Protection domain.

Claims (1)

1. a kind of multiple operating modes process modal identification method, which is characterized in that specifically comprise the following steps：
Step 1：Off-line training is as follows：
Step 1.1：The historical data under d kind difference operating modes is collected, training dataset is established
Wherein, m is number of probes, niFor the number of samples under i-th of operating mode；
Step 1.2：Calculate the priori weight beta of each operating modei：
Step 1.3：Data under i-th of operating mode are pre-processed：
Wherein,Represent the measurement vector of k-th of sampling instant under i-th of operating mode；For corresponding zero averaging to Amount；μ{i}For the sample average of i-th of floor data, calculated by formula (3)：
Step 1.4：Calculate the sample covariance Q of i-th of floor data{i}：
Wherein,ForTransposition；
Step 1.5：Enable j=1, and step-up error threshold value ∈；
Step 1.6：K=0 is enabled, random vector is generatedAnd it normalizes：
Wherein,It indicates2 norms；
Step 1.7：If j>1, then in formula (5)It is orthogonalized processing：
Wherein, wiIt is the i-th row of orthogonal matrix W to be optimized；Middle k is the kth step of iterative calculation, and T is transposition；
Step 1.8：Calculate function g (wj)：
Wherein, wjIt is the jth row of orthogonal matrix W to be optimized；
Step 1.9：Calculate g (wj) Jacobian matrix J (wj)：
Wherein, I is unit battle array；
Step 1.10：Newton iteration：
Step 1.11：If j>1, in formula 9It is orthogonalized processing：
Step 1.12：To in formula 10It is normalized：
Step 1.13：In judgment formula (11), ifK=k+1 is enabled, and jumps to step 1.10 and carries out newton Iteration；
Step 1.14：J=j+1 is enabled, if j≤m, jumps to step 1.6；
Step 1.15：Orthogonal matrix W is obtained, the data of different operating modes are handled as follows：
z{i}=WTx{i}(12)；
Wherein, z{i}For the data vector under i-th of operating mode；
Step 1.16：Kernel function K and bandwidth h is set；
Step 1.17：To i=1 ..., d, j=1 ..., m estimate conditional probability density function：
Wherein,Indicate z under i-th of operating mode{i}The value of k-th of sampling instant of j-th of sensor；
Step 2：Online modal identification
Step 2.1：Following orthogonal transformation is carried out to sample data：
zk=WTxk(14)；
Step 2.2：To i=1 ..., d, design conditions probability density
Step 2.3：Mode classification is determined by maximal condition probability density：
CN201810207588.6A 2018-03-14 2018-03-14 Multi-working-condition process modal identification method Active CN108573278B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810207588.6A CN108573278B (en) 2018-03-14 2018-03-14 Multi-working-condition process modal identification method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810207588.6A CN108573278B (en) 2018-03-14 2018-03-14 Multi-working-condition process modal identification method

Publications (2)

Publication Number Publication Date
CN108573278A true CN108573278A (en) 2018-09-25
CN108573278B CN108573278B (en) 2021-06-22

Family

ID=63573968

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810207588.6A Active CN108573278B (en) 2018-03-14 2018-03-14 Multi-working-condition process modal identification method

Country Status (1)

Country Link
CN (1) CN108573278B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020180586A1 (en) * 2001-05-30 2002-12-05 Kitson Frederick Lee Face and environment sensing watch
US20030018459A1 (en) * 2001-06-20 2003-01-23 O' Riordan Donald J. Method for debugging of analog and mixed-signal behavioral models during simulation
CN104019799A (en) * 2014-05-23 2014-09-03 北京信息科技大学 Relative orientation method by using optimization of local parameter to calculate basis matrix
CN105486474A (en) * 2015-11-30 2016-04-13 上海卫星工程研究所 Satellite flexible part on-orbit modal identification realization system and method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020180586A1 (en) * 2001-05-30 2002-12-05 Kitson Frederick Lee Face and environment sensing watch
US20030018459A1 (en) * 2001-06-20 2003-01-23 O' Riordan Donald J. Method for debugging of analog and mixed-signal behavioral models during simulation
CN104019799A (en) * 2014-05-23 2014-09-03 北京信息科技大学 Relative orientation method by using optimization of local parameter to calculate basis matrix
CN105486474A (en) * 2015-11-30 2016-04-13 上海卫星工程研究所 Satellite flexible part on-orbit modal identification realization system and method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
LIANG CAO ET AL.: "Integrated Fault/State Estimation for Two-Dimensional Linear Time-Varying Systems", 《2017 CHINESE AUTOMATION CONGRESS(CAC)》 *
卢春红 等.: "基于贝叶斯推理的PKPCAM的非线性多模态过程故障检测与诊断方法", 《化工学报》 *

Also Published As

Publication number Publication date
CN108573278B (en) 2021-06-22

Similar Documents

Publication Publication Date Title
Chen et al. Graph convolutional network-based method for fault diagnosis using a hybrid of measurement and prior knowledge
Cai et al. A new fault detection method for non-Gaussian process based on robust independent component analysis
CN111401460B (en) Abnormal electric quantity data identification method based on limit value learning
Dirvanauskas et al. Embryo development stage prediction algorithm for automated time lapse incubators
CN103544499B (en) The textural characteristics dimension reduction method that a kind of surface blemish based on machine vision is detected
CN105472733B (en) It is a kind of applied in indoor positioning based on AP selective positioning method
Zhang et al. Batch process fault detection and identification based on discriminant global preserving kernel slow feature analysis
CN110880024B (en) Nonlinear process fault identification method and system based on discrimination kernel slow characteristic analysis
Cheng et al. Estimating the shift size in the process mean with support vector regression and neural networks
Dong et al. Joint data-driven fault diagnosis integrating causality graph with statistical process monitoring for complex industrial processes
Gu et al. Identification of concurrent control chart patterns with singular spectrum analysis and learning vector quantization
Zhang et al. Process monitoring, fault diagnosis and quality prediction methods based on the multivariate statistical techniques
Jang et al. Parameter estimation in nonlinear chemical and biological processes with unmeasured variables from small data sets
CN110046377A (en) A kind of selective ensemble instant learning soft-measuring modeling method based on isomery similarity
CN111914887A (en) Novel multi-mode chemical process abnormal state detection method
CN114879628A (en) Multi-mode industrial process fault diagnosis method based on local maximum mean difference resistance
CN109325065A (en) Multi-sampling rate flexible measurement method based on dynamic latent variable model
CN111913415B (en) Continuous stirring reaction kettle operation state monitoring method based on time sequence data analysis
CN110084301A (en) A kind of multiple operating modes process industry and mining city method based on hidden Markov model
CN108573278A (en) A kind of multiple operating modes process modal identification method
CN106599391B (en) Association vector machine soft measurement modeling method based on dynamic weighting of triangle angle values
Zhang et al. An ELM based online soft sensing approach for alumina concentration detection
Zhou et al. GD-RDA: a new regularized discriminant analysis for high-dimensional data
CN109598283A (en) A kind of aluminium electroloysis degree of superheat recognition methods based on semi-supervised extreme learning machine
CN103065029A (en) Gene selection method for tumor detection

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