CN109634240B - Dynamic process monitoring method based on novel dynamic principal component analysis - Google Patents
Dynamic process monitoring method based on novel dynamic principal component analysis Download PDFInfo
- Publication number
- CN109634240B CN109634240B CN201811577508.2A CN201811577508A CN109634240B CN 109634240 B CN109634240 B CN 109634240B CN 201811577508 A CN201811577508 A CN 201811577508A CN 109634240 B CN109634240 B CN 109634240B
- Authority
- CN
- China
- Prior art keywords
- matrix
- vector
- formula
- vectors
- principal component
- 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.)
- Active
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
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS], computer integrated manufacturing [CIM]
- G05B19/41885—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS], computer integrated manufacturing [CIM] characterised by modeling, simulation of the manufacturing system
-
- 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
- G05B2219/30—Nc systems
- G05B2219/32—Operator till task planning
- G05B2219/32339—Object oriented modeling, design, analysis, implementation, simulation language
-
- 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
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Abstract
The invention discloses a dynamic process monitoring method based on novel dynamic principal component analysis, which aims to simultaneously dig out variance characteristics and autocorrelation characteristics in training data and implement dynamic process monitoring on the basis of the variance characteristics and the autocorrelation characteristics. Specifically, the method redefines the objective function of the traditional principal component analysis algorithm, so that the objective function of the novel principal component analysis algorithm related in the invention simultaneously covers the variance characteristic and the autocorrelation characteristic. The method simultaneously considers the variance characteristic of the data and the autocorrelation characteristic on the time sequence during data characteristic mining. Thus. The novel principal component analysis algorithm related in the method is not only a brand-new feature extraction algorithm, but also can mine more comprehensive features. It can be said that the method of the present invention is a more preferred dynamic process monitoring method.
Description
Technical Field
The invention relates to a data-driven process monitoring method, in particular to a dynamic process monitoring method based on novel dynamic principal component analysis.
Background
In recent years, in the industrial 'big data' hot tide, modern industrial processes gradually move towards digital management. The method is mainly benefited from the rapid development and wide application of the advanced instrument technology and the computing technology, and the production process object can store mass data off line and measure mass data on line. The data contains information which can reflect the running state of the production process, and the monitoring of the running state of the process by utilizing the sampling data is favored by more scholars. In the last decade, both academic and industrial fields, a great deal of manpower and material resources have been invested in the research of process monitoring methods with fault detection and diagnosis as core tasks. In the field of data-driven process monitoring research, statistical process monitoring is the most studied method, and the Principal Component Analysis (PCA) algorithm is the most popular implementation technique. The core essence of statistical process monitoring is that: and (3) carrying out feature mining on the sampled data under the normal working condition, and establishing a single-classification model by using the features to carry out monitoring. Therefore, feature mining or feature extraction plays an important role in the field of statistical process monitoring.
When a data-driven process monitoring model is established, the characteristics of the sampled data under normal working conditions need to be fully mined. Due to the short sampling time interval, the sampled data inevitably has autocorrelation in time sequence. Thus, the dynamic nature of the data autocorrelation is a problem that must be considered. For the research of the dynamic process monitoring problem, the most common idea is to use an augmentation matrix, confuse the autocorrelation and cross correlation of data, and then use the PCA algorithm to perform feature extraction. The dynamic PCA method is the most classical dynamic PCA method in scientific research documents, and on the basis of the classical dynamic PCA method, a plurality of novel modeling ideas are invented by scholars, so that the dynamic process monitoring performance of the dynamic PCA method is improved. However, the objective function of the PCA algorithm determines that the PCA model can only mine variance information in the training data, and cannot mine the correlation in time series. Of course, both the variance characteristic and the autocorrelation characteristic of normal data are potential characteristics of data, and one of the characteristics cannot be emphasized unilaterally.
Disclosure of Invention
The invention aims to solve the main technical problems that: and simultaneously excavating variance characteristics and autocorrelation characteristics in the training data, and carrying out dynamic process monitoring on the basis of the variance characteristics and the autocorrelation characteristics. Therefore, the invention discloses a novel dynamic principal component analysis algorithm which is applied to dynamic process monitoring. Specifically, the method redefines the objective function of the traditional PCA algorithm, so that the objective function simultaneously covers the variance characteristic and the autocorrelation characteristic.
The technical scheme adopted by the invention for solving the technical problems is as follows: a dynamic process monitoring method based on novel dynamic pivot analysis comprises the following steps:
(1) collecting samples under normal operation condition in production process to form a training data matrix X belonging to Rn×mAnd calculating the mean value mu of each column vector in the matrix X1,μ2,…,μmAnd standard deviation delta1,δ2,…,δmCorresponding component mean vector μ ═ μ1,μ2,…,μm]TAnd the standard deviation vector delta ═ delta1,δ2,…,δm]Wherein n is the number of training samples, m is the number of process measurement variables, R is the set of real numbers, R is the number of training samplesn×mRepresenting a matrix of real numbers in dimensions n x m, the upper index T representing the transpose of the matrix or vector.
(2) According to the formulaNormalizing the matrix X to obtain a matrixWherein U is E.Rn×mIs a matrix consisting of n identical mean vectors μ, i.e. U ═ μ, μ, …, μ]TThe diagonal elements in the diagonal matrix Φ consist of the standard deviation vector δ.
(3) Memory matrixAfter setting the autocorrelation order to D, according to the formula Xd=[xd,xd+1,…,xn-D+d-1]TConstruction matrix X1,X2,…,XD+1Wherein the subscript D ═ 1, 2, …, D +1, xiFor the normalized ith data sample, i is 1, 2, …, n.
To be provided withFor training data, the objective function of the conventional PCA algorithm is as follows:
in the above formula, the load vector p ∈ Rm×1Is the solution of the objective function of the PCA algorithm. As is apparent from equation (1), the PCA algorithm is aimed at maximizing the variance of the feature components when mining the latent features, and does not consider the correlation features in the time series.
On the basis of formula (1), the method defines an objective function of a novel principal component analysis algorithm as follows:
in the above formula, the vector β ═ β1,β2,…,βD]T,The Kronecker product is expressed, and the specific calculation mode is as follows:
from the comparison between formula (2) and formula (1), it can be seen that the objective function defined by the method of the present invention is added with a term on the basis of the PCA algorithm objective function:this term is primarily reflected in the correlation over time series. Therefore, when the load vector p is solved, the variance characteristic and the autocorrelation characteristic in the training data are considered, and more useful characteristics can be excavated compared with the traditional PCA algorithm.
The target function of the method can be solved by a Lagrange multiplier method, and a Lagrange function J shown as follows is firstly designed by utilizing a Lagrange multiplier lambda and a gamma:
next, the partial derivative of the function J with respect to the vector p and the vector β is calculated:
forcing the partial derivatives in equations (5) and (6) to be equal to zero yields the following equation:
it can be seen that equation (7) describes a eigenvalue problem and equation (8) describes the relationship between vector p and vector β. The load vector p is a feature vector corresponding to the maximum feature value of the feature value problem in equation (7). Since the vector β is needed for solving the eigenvalue problem in the formula (7) and the vector p is needed for solving the vector β in the formula (8), the vector p and the vector β can be solved through mutual iteration of the formula (7) and the formula (8) until convergence.
(4) After the number of the load vectors is set to be K, K load vectors p are solved by utilizing a novel principal component analysis algorithm1,p2,…,pKThe specific implementation procedure is as follows.
(4.1) initialization vector β ═ 1, 1, …, 1]TI.e. all elements in the vector beta are equal to 1.
(4.2) processing the vector β according to the formula β/| β | | | normalization, where the notation | | β | | | represents the length of the calculation vector β.
(4.3) calculating the eigenvector p corresponding to the maximum eigenvalue in the formula (7)kThis step requires a feature vector pkIs a unit length.
(4.5) determine whether the vector β converges? The judgment standard is as follows: until each element in the vector beta is not changed any more, if not, returning to the step (4.2); if yes, a k load vector p is obtainedk。
(4.7) according to the formulaUpdating a matrixAnd according to formula Xd=[xd,xd+1,…,xn-D+d-1]TConstruction matrix X1,X2,…,XD+1。
(4.8) repeating the steps (4.1) to (4.7) until K load vectors p are obtained1,p2,…,pKK regression vectors w1,w2,…,wKAnd K score vectors t1,t2,…,tK。
(5) According to the formula Θ ═ P (W)TP)-1Computing a projective transformation matrix Θ, wherein the matrix W ═ W1,w2,…,wK]The load matrix P ═ P1,p2,…,pK]。
(6) Calculating a scoring matrix T ═ T1,t2,…,tK]Is given by the covariance matrix Λ ═ TTT/(n-1), and calculates the control limit ψ as followslimAnd Qlim:
In the above two formulae, FK,n-K,αRepresenting the confidence coefficient as alpha and the degree of freedom as the corresponding values of the F distribution of K and n-K respectively,the degree of freedom is h, the confidence coefficient is alpha, a and tau are respectivelyRespectively, the estimated mean and the estimated variance of the Q statistic.
The steps (1) to (6) are the off-line modeling stage of the method of the present invention, and the mean and standard deviation vectors in the step (1), the projection transformation matrix Θ and the load matrix P in the step (5), and the covariance matrix Λ and the upper control limit in the step (6) need to be retained for the following on-line monitoring process invocation.
(7) Collecting data sample x ∈ R at new sampling momentm×1According to the formulaNormalizing x to obtain
(8) The specific values of statistics ψ and Q are calculated according to the formula shown below:
(9) judging whether the condition psi is less than or equal to psilimAnd Q is less than or equal to QlimIs there a If yes, the current sample is collected from a normal working condition, and the step (7) is returned to continue monitoring the sample data at the next moment; if not, the current monitoring sample is collected from the fault working condition.
Compared with the traditional method, the method has the advantages that:
in the process of extracting the potential features, the method simultaneously considers the variance features of the data and the autocorrelation features on the time sequence. Thus. The novel principal component analysis algorithm related in the method can mine more comprehensive characteristics. Although the novel principal component analysis algorithm involved in the method is improved based on the classical PCA algorithm, the novel principal component analysis algorithm is a brand-new feature extraction algorithm.
Drawings
FIG. 1 is a flow chart of an embodiment of the method of the present invention.
FIG. 2 is a flow chart of the implementation of the novel principal component analysis algorithm in the method of the present invention.
Detailed Description
The method of the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.
As shown in fig. 1, the present invention discloses a dynamic process monitoring method based on a novel dynamic principal component analysis, and the specific implementation manner is as follows.
Step (1): collecting samples under normal operation condition in production process to form a training data matrix X belonging to Rn×mAnd calculating the mean value mu of each column vector in the matrix X1,μ2,…,μmAnd standard deviation delta1,δ2,…,δmCorresponding component mean vector μ ═ μ1,μ2,…,μm]TAnd the standard deviation vector delta ═ delta1,δ2,…,δm]。
Step (2): according to the formulaNormalizing the matrix X to obtain a matrixWherein U is E.Rn×mIs a matrix consisting of n identical mean vectors μ, i.e. U ═ μ, μ, …, μ]TThe diagonal elements in the diagonal matrix Φ consist of the standard deviation vector δ.
And (3): memory matrixAfter setting the autocorrelation order to D, according to the formula Xd=[xd,xd+1,…,xn-D+d-1]TConstruction matrix X1,X2,…,XD+1Wherein the index D is 1, 2, …, D + 1.
And (4): after the number of the load vectors is set to be K, K load vectors p are solved by utilizing a novel principal component analysis algorithm1,p2,…,pK. As shown in fig. 2, a flowchart of an implementation of the novel principal component analysis algorithm according to the present invention includes steps (4.1) to (4.8).
Step (4.1): initialization vector β ═ 1, 1, …, 1]TI.e. all elements in the vector beta are equal to 1.
Step (4.2): the vector β is processed normalized according to the formula β/| β | |, where the notation | | β | | | represents the length of the calculation vector β.
Step (4.3): calculating the eigenvector p corresponding to the largest eigenvalue in the eigenvalue problem as shown belowkThis step requires a feature vector pkIs a unit length:
in the above formula, λ is a feature value, and p is a feature vector.
Step (4.5): determine whether vector β converges? If not, returning to the step (4.2); if yes, a k load vector p is obtainedk。
Step (4.6): according to the formulaAndcalculating score vectors t respectivelykAnd the regression vector wk。
Step (4.7): according to the formulaUpdating a matrixAnd according to formula Xd=[xd,xd+1,…,xn-D+d-1]TConstruction matrix X1,X2,…,XD+1。
Step (4.8): repeating the steps (4.1) to (4.7) until K load vectors p are obtained1,p2,…,pKK regression vectors w1,w2,…,wKAnd K score vectors t1,t2,…,tK。
And (5): according to the formula Θ ═ P (W)TP)-1Computing a projective transformation matrix Θ, wherein the matrix W ═ W1,w2,…,wK]The load matrix P ═ P1,p2,…,pK];
And (6): calculating a scoring matrix T ═ T1,t2,…,tK]Is given by the covariance matrix Λ ═ TTT/(n-1), and calculates the control limit ψ as followslimAnd Qlim:
In the above two formulae, FK,n-K,αRepresenting the confidence coefficient as alpha and the degree of freedom as the corresponding values of the F distribution of K and n-K respectively,representing the degree of freedom h and the confidence coefficient alpha as the corresponding values of chi-square distribution, and a and tau are the estimated mean value and the estimated variance of the Q statistic respectively;
and (7): collecting data sample x ∈ R at new sampling momentm×1According to the formulaNormalizing x to obtain
And (8): the specific values of statistics ψ and Q are calculated according to the formula shown below:
and (9): judging whether the condition psi is less than or equal to psilimAnd Q is less than or equal to QlimIs there a If yes, the current sample is collected from a normal working condition, and the step (7) is returned to continue monitoring the sample data at the next moment; if not, the current monitoring sample is collected from the fault working condition.
The above embodiments are merely illustrative of specific implementations of the present invention and are not intended to limit the present invention. Any modification of the present invention within the spirit of the present invention and the scope of the claims will fall within the scope of the present invention.
Claims (1)
1. A dynamic process monitoring method based on novel dynamic pivot analysis is characterized by comprising the following steps:
step (1): collecting samples under normal operation condition in production process to form a training data matrix X belonging to Rn×mAnd calculating the mean value mu of each column vector in the matrix X1,μ2,…,μmAnd standard deviation delta1,δ2,…,δmCorresponding component mean vector μ ═ μ1,μ2,…,μm]TAnd the standard deviation vector delta ═ delta1,δ2,…,δm]Wherein n is the number of training samples, m is the number of process measurement variables,r is a real number set, Rn×mA real number matrix of n × m dimensions is represented, and the upper label T represents the transpose of a matrix or a vector;
step (2): according to the formulaNormalizing the matrix X to obtain a matrixWherein U is E.Rn×mIs a matrix consisting of n identical mean vectors μ, i.e. U ═ μ, μ, …, μ]TThe elements on the diagonal in the diagonal matrix phi consist of a standard deviation vector delta;
and (3): memory matrixAfter setting the autocorrelation order to D, according to the formula Xd=[xd,xd+1,…,xn-D+d-1]TConstruction matrix X1,X2,…,XD+1Wherein the subscript D ═ 1, 2, …, D +1, xiThe normalized ith data sample is i ═ 1, 2, …, n;
and (4): setting the number of load vectors as K, and solving K load vectors p by using a novel principal component analysis algorithm1,p2,…,pKThe specific implementation process is shown as follows;
step (4.1): initialization vector β ═ 1, 1, …, 1]TI.e. all elements in the vector β are equal to 1;
step (4.2): processing the vector beta according to a formula beta/beta normalized, wherein the symbol beta represents the length of the calculation vector beta;
step (4.3): calculating the eigenvector p corresponding to the largest eigenvalue in the eigenvalue problem as shown belowkThis step requires a feature vector pkIs a unit length:
in the above formula, λ is a characteristic value, and p is a characteristic vector;
step (4.5): judging whether the vector beta is converged; if not, returning to the step (4.2); if yes, a k load vector p is obtainedk;
Step (4.6): according to the formulaCalculating score vectors t respectivelykAnd the regression vector wk;
Step (4.7): according to the formulaUpdating a matrixAnd according to formula Xd=[xd,xd+1,…,xn-D+d-1]TConstruction matrix X1,X2,…,XD+1;
Step (4.8): repeating the steps (4.1) to (4.7) until K load vectors p are obtained1,p2,…,pKK regression vectors w1,w2,…,wKAnd K score vectors t1,t2,…,tK;
And (5): according to the formula Θ ═ P (W)TP)-1Computing a projective transformation matrix Θ, wherein the matrix W ═ W1,w2,…,wK]The load matrix P ═ P1,p2,…,pK];
And (6): calculating a scoring matrix T ═ T1,t2,…,tK]Is given by the covariance matrix Λ ═TTT/(n-1), and calculating the control limit psi according to the formulalimAnd Qlim:
In the above two formulae, FK,n-K,αRepresenting the confidence coefficient as alpha and the degree of freedom as the corresponding values of the F distribution of K and n-K respectively,representing the degree of freedom h and the confidence coefficient alpha as the corresponding values of chi-square distribution, and a and tau are the estimated mean value and the estimated variance of the Q statistic respectively;
the steps (1) to (6) are an off-line modeling stage, and the mean value and standard deviation vector in the step (1), the projection transformation matrix Θ and the load matrix P in the step (5), and the covariance matrix Λ and the control upper limit in the step (6) need to be reserved for calling the on-line monitoring process shown in the following;
and (7): collecting data sample x ∈ R at new sampling momentm×1According to the formulaNormalizing x to obtain
And (8): the specific values of statistics ψ and Q are calculated according to the formula shown below:
and (9): judging whether the condition psi is less than or equal to psilimAnd Q is less than or equal to Qlim(ii) a If yes, the current sample is collected from a normal working condition, and the step (7) is returned to continue monitoring the sample data at the next moment; if not, the current monitoring sample is collected from the fault working condition.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811577508.2A CN109634240B (en) | 2018-12-13 | 2018-12-13 | Dynamic process monitoring method based on novel dynamic principal component analysis |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811577508.2A CN109634240B (en) | 2018-12-13 | 2018-12-13 | Dynamic process monitoring method based on novel dynamic principal component analysis |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109634240A CN109634240A (en) | 2019-04-16 |
CN109634240B true CN109634240B (en) | 2021-03-09 |
Family
ID=66076616
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811577508.2A Active CN109634240B (en) | 2018-12-13 | 2018-12-13 | Dynamic process monitoring method based on novel dynamic principal component analysis |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109634240B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111610021B (en) * | 2020-06-04 | 2022-05-10 | 沈阳科网通信息技术有限公司 | Early fault detection method for gearbox |
CN111913415B (en) * | 2020-06-13 | 2022-03-18 | 宁波大学 | Continuous stirring reaction kettle operation state monitoring method based on time sequence data analysis |
CN112229212A (en) * | 2020-08-18 | 2021-01-15 | 广东工业大学 | Roller kiln energy consumption abnormity detection method based on dynamic principal component analysis |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103631145A (en) * | 2013-12-11 | 2014-03-12 | 清华大学 | Monitoring index switching based multi-operating-mode process monitoring method and system |
CN105955219A (en) * | 2016-05-30 | 2016-09-21 | 宁波大学 | Distributed dynamic process fault detection method based on mutual information |
CN108388234A (en) * | 2018-03-19 | 2018-08-10 | 宁波大学 | A kind of fault monitoring method dividing changeable gauge block pca model based on correlation |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7447609B2 (en) * | 2003-12-31 | 2008-11-04 | Honeywell International Inc. | Principal component analysis based fault classification |
-
2018
- 2018-12-13 CN CN201811577508.2A patent/CN109634240B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103631145A (en) * | 2013-12-11 | 2014-03-12 | 清华大学 | Monitoring index switching based multi-operating-mode process monitoring method and system |
CN105955219A (en) * | 2016-05-30 | 2016-09-21 | 宁波大学 | Distributed dynamic process fault detection method based on mutual information |
CN108388234A (en) * | 2018-03-19 | 2018-08-10 | 宁波大学 | A kind of fault monitoring method dividing changeable gauge block pca model based on correlation |
Also Published As
Publication number | Publication date |
---|---|
CN109634240A (en) | 2019-04-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109634240B (en) | Dynamic process monitoring method based on novel dynamic principal component analysis | |
WO2022141669A1 (en) | Bearing fault diagnosis method for dynamic joint distribution alignment network under variable working conditions | |
CN107292023B (en) | Bridge structure state diagnosis method based on narrow-area characteristics of damage index system | |
CN109669415B (en) | Dynamic process monitoring method based on structured typical variable analysis | |
CN110084221B (en) | Serialized human face key point detection method with relay supervision based on deep learning | |
CN111142501B (en) | Fault detection method based on semi-supervised autoregressive dynamic hidden variable model | |
CN107065842B (en) | A kind of fault detection method based on particle group optimizing core independent component analysis model | |
CN109522972B (en) | Dynamic process monitoring method based on latent variable autoregressive model | |
CN109407649B (en) | Fault type matching method based on fault characteristic variable selection | |
CN108445867B (en) | non-Gaussian process monitoring method based on distributed ICR model | |
CN108469805B (en) | Distributed dynamic process monitoring method based on dynamic optimal selection | |
CN109542070B (en) | Dynamic process monitoring method based on dual-target optimization algorithm | |
CN108171271B (en) | Early warning method and system for equipment degradation | |
CN111913460A (en) | Fault monitoring method based on sequence correlation local preserving projection algorithm | |
CN111915121B (en) | Chemical process fault detection method based on generalized typical variable analysis | |
CN109669412B (en) | non-Gaussian process monitoring method based on novel dynamic independent component analysis | |
CN111914384B (en) | Chemical process state monitoring method based on autocorrelation latent variable model | |
CN109669414B (en) | Dynamic process monitoring method based on autocorrelation characteristic decomposition | |
CN109542974B (en) | Dynamic process monitoring method based on nonlinear dynamic component analysis | |
CN111913415B (en) | Continuous stirring reaction kettle operation state monitoring method based on time sequence data analysis | |
CN115778372A (en) | Knee joint angle estimation method based on surface electromyogram signals | |
CN114841063A (en) | Aero-engine residual life prediction method based on deep learning | |
CN111913462B (en) | Chemical fault monitoring method based on generalized multiple independent element analysis model | |
CN110288724A (en) | A kind of batch process monitoring method based on wavelet function pivot analysis | |
CN113159082A (en) | Incremental learning target detection network model construction and weight updating method |
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 |