CN113076211B - Quality-related fault diagnosis and false alarm feedback method based on fault reconstruction - Google Patents
Quality-related fault diagnosis and false alarm feedback method based on fault reconstruction Download PDFInfo
- Publication number
- CN113076211B CN113076211B CN202110333398.0A CN202110333398A CN113076211B CN 113076211 B CN113076211 B CN 113076211B CN 202110333398 A CN202110333398 A CN 202110333398A CN 113076211 B CN113076211 B CN 113076211B
- Authority
- CN
- China
- Prior art keywords
- fault
- quality
- representing
- matrix
- subspace
- 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
- 238000000034 method Methods 0.000 title claims abstract description 81
- 238000003745 diagnosis Methods 0.000 title claims abstract description 23
- 238000012544 monitoring process Methods 0.000 claims abstract description 33
- 238000004519 manufacturing process Methods 0.000 claims abstract description 21
- 238000012360 testing method Methods 0.000 claims abstract description 7
- 239000011159 matrix material Substances 0.000 claims description 123
- 239000013598 vector Substances 0.000 claims description 32
- 238000000354 decomposition reaction Methods 0.000 claims description 18
- 238000010276 construction Methods 0.000 claims description 9
- 230000007423 decrease Effects 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 6
- 238000012216 screening Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000000513 principal component analysis Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 230000017105 transposition Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 description 12
- 238000010586 diagram Methods 0.000 description 4
- 238000012549 training Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000006227 byproduct Substances 0.000 description 1
- 239000003054 catalyst Substances 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000007795 chemical reaction product Substances 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000000376 reactant Substances 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
- G06F11/0703—Error or fault processing not based on redundancy, i.e. by taking additional measures to deal with the error or fault not making use of redundancy in operation, in hardware, or in data representation
- G06F11/079—Root cause analysis, i.e. error or fault diagnosis
-
- 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
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- 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
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Biomedical Technology (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Complex Calculations (AREA)
Abstract
The invention discloses a quality-related fault diagnosis and false alarm feedback method based on fault reconstruction, which comprises the following steps: establishing an improved latent structure model; constructing a judging standard of a fault state in an industrial process based on the improved latent structure model; establishing a fault library according to the judging standard of the fault state; based on a fault library, carrying out fault identification and false alarm feedback on a monitored fault sample; and extracting a fault subspace of the unrecognizable monitoring fault sample, and updating a fault library according to the fault subspace. According to the invention, the fault type of the fault sample is identified based on fault reconstruction by constructing the quality-related and quality-independent fault library, and the false alarm sample is fed back, so that the online fault diagnosis and false alarm feedback of the test sample are realized.
Description
Technical Field
The invention belongs to the technical field of reliability engineering, and particularly relates to a quality-related fault diagnosis and false alarm feedback method based on fault reconstruction.
Background
With the development of modern technology, the more complex the industrial process system, the more cannot be described by a simple physical model and a mechanism model, so that a series of fault diagnosis methods based on fault trees and physical models are difficult to apply. With the development of integrated circuits, a large number of sensors are deployed in many critical locations in an industrial process, while obtaining a vast amount of data. Therefore, the data-driven modeling-based method is gradually becoming a research hotspot and is greatly developed and applied. In the data obtained by the industrial process, the characteristics of high coupling, non-Gaussian, non-linearity and the like are usually adopted, and how to propose a characteristic modeling of the process from the data is a difficult point of a data driving method. Meanwhile, when the performance of industrial equipment is degraded, parameters in the production flow drift and even a certain link fails, abnormal changes can occur to process data monitored by corresponding sensors, and how to find out the reasons of the failures from the changes is the key point and the difficulty of the current research.
In fault diagnosis, two types of fault forms can be classified. The sensor faults are fault phenomena caused by faults of the sensor or abnormal corresponding parts, and the sensor for monitoring the abnormality is usually positioned by adopting a contribution graph method so as to replace or remove the obstacle. The other type of faults are process faults, are fixed fault types formed by abnormal multi-sensor monitoring caused by specific reasons, and have the characteristics of obvious fault expression distinction and various types. For process faults, the form is more complex than sensor faults. In the diagnosis and identification of the fault type aiming at the key performance index, the fault type is influenced by quality irrelevant faults, false alarm is easy to occur, and the fault cause cannot be identified, so that a series of serious production problems are caused. To this end, two problems need to be solved: firstly, aiming at the identification of the fault type of the key performance index; and secondly, real-time feedback is carried out aiming at the fault of false alarm. In an actual production process, one type of process fault typically has a specific fault signature, which may be represented by a fault subspace. An accurate fault subspace can reflect all characteristics of faults through smaller dimensions, so that real-time efficiency in fault identification is further improved, and diagnosis is more accurate. The invention realizes the effective identification of quality related faults by efficiently extracting the quality related fault subspace. Meanwhile, a quality irrelevant fault subspace extraction method is designed, so that false alarm faults can be diagnosed in real time and feedback can be given in time.
Disclosure of Invention
Aiming at the defects in the prior art, the quality-related fault diagnosis and false alarm feedback method based on fault reconstruction solves the problems in the prior art.
In order to achieve the aim of the invention, the invention adopts the following technical scheme: a quality-related fault diagnosis and false alarm feedback method based on fault reconstruction comprises the following steps:
s1, establishing an improved latent structure model;
s2, constructing a judging standard of a fault state in an industrial process based on the improved latent structure model;
s3, establishing a fault library according to the judging standard of the fault state;
s4, based on a fault library, carrying out fault identification and false alarm feedback on the monitored fault sample;
and S5, extracting a fault subspace of the unrecognizable monitoring fault sample, and updating a fault library according to the fault subspace.
Further, the step S1 specifically includes:
s1.1, collecting a plurality of input data X and output data Y in an industrial process, wherein a column of the input data X comprises m variables, and a row of the input data X comprises n sampling samples; the column of output data Y includes p variables, the row of which includes n sample samples;
s1.2, setting the iteration number i=1;
s1.3, randomly fetching a column of data u of the output data Y i Let the first intermediate coefficient u old ＝u i Data u i The updating is carried out specifically as follows:
|w i |＝w i /norm(w i )
t i ＝X i |w i |/|w i | T |w i |
q i ＝Y i t i /t i T t i
u i '＝Y i q i /q i T q i
wherein u is i ' represents updated data u i ，Y i Representing the output Y, q after the ith update i Representing coefficient vectors, T representing transpose, X i Representing after the ith updateX of (2) i ，t i Representing data X i Along the modulus |w i Dimension reduction data after projection, |w i The I represents the regression vector w i Norm (x) represents the operation of taking the binary norm;
s1.4, judging the I U old -u i '||/||u old ||<10 -4 If yes, entering a step S1.5, otherwise, adding one to the count value of i, and returning to the step S1.3;
s1.5 according to data X i Dimension t decreases i Obtaining a load vector p i The method comprises the following steps:
p i ＝X i T t i /t i T t i
s1.6 according to the load vector p i Coefficient vector q i Dimension t decreases i Data X i Data Y i The updating is carried out specifically as follows:
X i+1 ＝X i -t i p i T
Y i+1 ＝Y i -t i q i T
wherein X is i+1 Representing updated data X i ，Y i+1 Representing updated Y i '；
S1.7, judging whether the iteration times i are equal to A, if so, entering a step S1.8, otherwise, adding one to the count value of i, and returning to the step S1.3;
s1.8, the A-time updated reduced data is formed into a scoring matrix T= (T) 1 ,...,t A ) The regression vector updated a times is formed into a weight matrix w= (W) 1 ,...,w A ) The load vector updated a times is formed into a load matrix p= (P) 1 ,...,p A )；
S1.9, according to the scoring matrix T, the weight matrix W and the load matrix P, establishing an improved latent structure model as follows:
wherein the method comprises the steps ofRepresenting quality-related subspaces, +.>Representing quality independent subspaces, T m Represented by X along P m Score matrix obtained by dimension reduction, P m Representing projection vectors, M representing a first coefficient matrix; />Representing a predictable space, which is information directly affected by X; />Representing unpredictable space, which is space not directly affected by X; r represents a second coefficient matrix, Q represents a load matrix of Y,>representing residual projection vector, Λ m Representing a matrix of eigenvalues.
Further, the step S2 specifically includes:
s2.1, based on the improved latent structure model, acquiring statistics of a single sample X in the input data X in a quality related subspace and a quality independent subspace as follows:
wherein D represents statistics of quality-related subspacesQuantity, T denotes transpose, T m Representing a score vector, Λ representing a second intermediate coefficient;
s2.2, constructing control limits of statistics:
wherein C is D Representing a control limit of a quality-related subspace, F A,n-A,α F distribution with degrees of freedom A and n-A and confidence alpha is represented;
s2.3, judging whether the statistic D of the quality related subspace is larger than the control limit C D If yes, judging that the sample x fails, otherwise, judging that the sample x does not fail;
s2.4, acquiring the fault state of each sample in the input data X in the industrial process according to the method in the steps S2.1-S2.3, and completing the construction flow of the judging standard of the fault state in the industrial process.
Further, establishing the fault library in the step S3 includes establishing a quality-related fault library and a quality-independent fault library.
Further, the specific steps of establishing the quality-related fault library are as follows:
s3.1.1 construction of quality-related fault datasets based on prior experienceWherein k is 1 For the total number of fault types>All represent fault data;
s3.1.2 fault data X f,1 Projection to a quality related subspace and a quality independent subspace, specifically:
wherein T is f,1 A scoring matrix representing the fault data,representing a load matrix->Representing a quality-related fault space resulting from projection of quality-related fault data into a quality-related subspace,/a quality-related fault space resulting from projection of quality-related fault data into a quality-related subspace>Representing a quality-related fault space obtained by projecting quality-related fault data into a quality-related subspace, wherein T represents transposition;
s3.1.3 calculating quality-related fault spaceAnd quality-related subspace->Specifically, the covariance matrix of (a) is:
wherein C is f,1 Representing quality-related fault spacesC represents the quality independent fault space +.>Is a covariance matrix of (a);
s3.1.4 extracting covariance matrix C by generalized principal component analysis f,1 Generalized eigenvector V of sum covariance matrix C f The method comprises the following steps:
V f ＝eig(C f,1 ,C)
wherein eig represents a eigenvalue decomposition operation;
s3.1.5 according to the generalized eigenvector V f Extracting quality-related fault spaceRemoving the normal information in the database to obtain a quality-related fault space after removing the normal information +.>The method comprises the following steps:
s3.1.6 to quality-related fault space after removing normal informationSingular value decomposition is carried out, specifically:
wherein U is f,1 A projection matrix representing a singular value decomposition,right singular value matrix representing singular value decomposition, D f,1 Representing the matrix of singular values on the diagonal in descending order;
s3.1.7 screening out projection matrix U f,1 The first L feature vectors make X f,1 Statistics D after sample reconstruction in (a) is smaller than control limit C D Constructing the first L eigenvectors as a failure subspace Σ f,1 ；
S3.1.8 according to the method of steps S3.1.2-S3.1.7, obtainingCorresponding failure subspace->Constructing a quality-related fault library as sigma f ＝(Σ f,1 ，...,Σ f,k )。
Further, the specific steps of establishing the quality independent fault library are as follows:
s3.2.1 based on improved latent structure model, for predictable spaceThe main component is decomposed, specifically:
wherein T is y Representing a predictable spaceScore matrix of->Representing +.>Projection matrix obtained by PCA decomposition +.>The number of principal elements of A y ，A y Rank (Q), rank (x) represents the rank calculation;
s3.2.2 according to the scoring matrix T y Obtaining a load matrix P y And according to the load matrix P y Score matrix T y ObtainingThe method comprises the following steps:
wherein ( -1 The operation of inversion is represented by the expression,representation->Is a complete mass space of (2);
s3.2.3 according to the load matrix P y Score matrix T y Acquisition ofQuality independent matrix->The method comprises the following steps:
s3.2.4 matrix independent of qualityThe main component is decomposed, specifically:
wherein T is o Representing quality independent matricesScore matrix, P o Representing a load matrix->The number of the medium pivot elements is A-A y ；
S3.2.5 according to the complete mass spaceQuality independent matrix->Optimizing the improved latent structure model to obtain an optimized latent structure model which is:
s3.2.6 construction of quality independent fault datasets based on prior experienceWherein k is 2 Representing the total number of quality independent fault types, +.>Failure data;
s3.2.7 based on the optimized latent structure model, fault data are obtainedProjection to a quality related subspace and a quality independent subspace, specifically:
wherein,representing quality-related subspace->Is>Representing fault data +.>Corresponding quality-related subspace, < >>Representing fault data +.>Corresponding quality independent matrix, < >>Representation->Score matrix of->Representing fault data +.>Corresponding scoring matrix, ">Representing fault data +.>A corresponding quality independent subspace;
s3.2.8 obtaining quality independent matrixAnd quality independent matrix->The covariance matrix of (2) is:
wherein,representing quality independent matrix->Corresponding covariance matrix, C represents quality independent matrix +.>A corresponding covariance matrix;
s3.2.9 extracting covariance matrix by generalized principal component algorithmGeneralized eigenvector of sum covariance matrix CThe method comprises the following steps:
s3.2.10 according to generalized eigenvectorExtracting quality independent matrix->The normal information in (a) is removed, specifically:
wherein,representing a fault space after the normal information is removed;
s3.2.11 to fault space after removing normal informationSingular value decomposition is carried out, specifically:
wherein,represents the (u) 1 Total fault subspace of individual quality independent faults, < ->A matrix representing a diagonal of descending singular value arrangement;
s3.2.12 screening out Total Fault subspacesMiddle and top L eigenvectors make +.>Statistics D after sample reconstruction in (a) is smaller than control limit C D The first L eigenvectors are built as failure subspace +.>
S3.2.13 according to the method of steps S3.2.6-S3.2.12, obtainingCorresponding failure subspace->Constructing a quality independent fault library as +.>
Further, the step S4 specifically includes:
s4.1, acquiring a monitoring fault sample x based on a fault database new In quality-related fault bank Σ f The fault amplitude f corresponding to each type of fault i The method comprises the following steps:
wherein i=1, 2,..k 1, ( + Represents the generalized inverse, T represents the transpose, Ω represents the third intermediate coefficient, x new Representing real-time data measured by m sensors in an on-line test and being a failure sample,the representation dimension is m×1;
s4.2, for monitoring fault sample x new Performing fault reconstruction, and reconstructing a sample x * The method comprises the following steps:
x * ＝x new -Σ f,i f i
s4.3, traversing the quality-related fault library Σ f Obtaining statistics D of the reconstructed samples and judging whether the statistics D of the reconstructed samples are smaller than or equal to a control limit C D If yes, identifying the fault as an i-th type fault, otherwise, entering a step S4.4;
s4.4, obtaining a monitoring fault sample x new In a quality independent fault bank Σ u The fault amplitude f corresponding to each type of fault i The method comprises the following steps:
s4.5, for monitoring fault sample x new Performing fault reconstruction, and reconstructing a sample x * The method comprises the following steps:
x * ＝x new -Σ ui f i
s4.4, traversal quality independenceFault bank Σ u Obtaining statistics D of the reconstructed samples and judging whether the statistics D of the reconstructed samples are smaller than or equal to a control limit C D If yes, the fault is identified as false alarm, false alarm feedback is carried out, otherwise, the fault is identified as unknown type fault, and the step S5 is carried out.
Further, the step S5 specifically includes:
s5.1, failure sample x is to be monitored new Corresponding output y new The input improved latent structure model is as follows:
wherein,representing a monitoring fault sample x new Quality-related subspace of->Representing a monitoring fault sample x new Quality independent subspace of->Representing the output y new Quality-related subspace of->Representing the output y new Quality independent subspaces of (2);
s5.2, acquiring quality independent subspacesStatistics Q of f,y The method comprises the following steps:
s5.3, obtaining control of each output sample y under normal working conditionsLimited C SPEy The method comprises the following steps:
where g 'represents a fourth intermediate coefficient, h' represents a fifth intermediate coefficient,the method is characterized in that chi-square distribution is represented, xi represents the variance of the historical statistics SPE under normal working conditions, and mu represents the mean value of the historical statistics SPE;
s5.4, judging statistic Q f,y Whether or not it is greater than the control limitIf yes, the unrecognizable monitoring fault sample is a quality related fault, otherwise, the unrecognizable monitoring fault sample is a quality unrelated fault.
The beneficial effects of the invention are as follows:
(1) The invention provides a quality-related fault diagnosis and false alarm feedback method based on fault reconstruction.
(2) The invention can rapidly identify faults in the industrial process, has high identification accuracy, improves identification efficiency and has wide implementation prospect.
Drawings
Fig. 1 is a flow chart of a quality-related fault diagnosis and false alarm feedback method based on fault reconstruction.
Fig. 2 is a diagram of IDV (8) fault reconstruction in an embodiment of the present invention.
Fig. 3 is a diagram of a fault reconstruction result of an IDV (1, 6, 8) fault by using a fault subspace of the IDV (2) fault in an embodiment of the present invention.
Fig. 4 is a diagram of a fault reconstruction result of the fault subspace corresponding to the IDV (1, 2, 13) fault in the embodiment of the present invention.
Fig. 5 is a graph of quality independent fault reconstruction results for IDV (3, 4,9, 11) faults in an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and all the inventions which make use of the inventive concept are protected by the spirit and scope of the present invention as defined and defined in the appended claims to those skilled in the art.
Embodiments of the present invention are described in detail below with reference to the accompanying drawings.
As shown in fig. 1, a quality-related fault diagnosis and false alarm feedback method based on fault reconstruction includes the following steps:
s1, establishing an improved latent structure model;
s2, constructing a judging standard of a fault state in an industrial process based on the improved latent structure model;
s3, establishing a fault library according to the judging standard of the fault state;
s4, based on a fault library, carrying out fault identification and false alarm feedback on the monitored fault sample;
and S5, extracting a fault subspace of the unrecognizable monitoring fault sample, and updating a fault library according to the fault subspace.
In this embodiment, X is a matrix of n rows and m columns of input data, i.e. m sensors are arranged to collect n samples; y is a matrix of output data, n rows and p columns, i.e. p sensors are arranged, and n samples are collected.
The step S1 specifically comprises the following steps:
s1.1, collecting a plurality of input data X and output data Y in an industrial process, wherein a column of the input data X comprises m variables, and a row of the input data X comprises n sampling samples; the column of output data Y includes p variables, the row of which includes n sample samples;
s1.2, setting the iteration number i=1;
s1.3, randomly fetching a column of data u of the output data Y i Let the first intermediate coefficient u old ＝u i Data u i The updating is carried out specifically as follows:
w i ＝X i T u i /u i T u i
|w i |＝w i /norm(w i )
t i ＝X i |w i |/|w i | T |w i |
q i ＝Y i t i /t i T t i
u i '＝Y i q i /q i T q i
wherein u is i ' represents updated data u i ，Y i Representing the output Y, q after the ith update i Representing coefficient vectors, T representing transpose, X i Representing X after the ith update i ，t i Representing data X i Along the modulus |w i Dimension reduction data after projection, |w i The I represents the regression vector w i Norm (x) represents the operation of taking the binary norm;
s1.4, judging the I U old -u i '||/||u old ||<10 -4 If yes, entering a step S1.5, otherwise, adding one to the count value of i, and returning to the step S1.3;
s1.5 according to data X i Dimension t decreases i Obtaining a load vector p i The method comprises the following steps:
p i ＝X i T t i /t i T t i
s1.6 according to the load vector p i Coefficient vector q i Dimension t decreases i Data X i Data Y i The updating is carried out specifically as follows:
X i+1 ＝X i -t i p i T
Y i+1 ＝Y i -t i q i T
wherein X is i+1 Representing updated data X i ，Y i+1 Representing updated Y i '；
S1.7, judging whether the iteration times i are equal to A, if so, entering a step S1.8, otherwise, adding one to the count value of i, and returning to the step S1.3;
s1.8, the A-time updated reduced data is formed into a scoring matrix T= (T) 1 ,...,t A ) The regression vector updated a times is formed into a weight matrix w= (W) 1 ,...,w A ) The load vector updated a times is formed into a load matrix p= (P) 1 ,...,p A )；
S1.9, according to the scoring matrix T, the weight matrix W and the load matrix P, establishing an improved latent structure model as follows:
wherein,representing quality-related subspaces, +.>Representing quality independent subspaces, T m Represented by X along P m Score matrix obtained by dimension reduction, P m Representing projection vectors, M representing a first coefficient matrix; />Representing a predictable space, which is directly X-bearingInformation of the influence; />Representing unpredictable space, which is space not directly affected by X; r represents a second coefficient matrix, Q represents a load matrix of Y,>representing residual projection vector, Λ m Representing a matrix of eigenvalues.
The step S2 specifically comprises the following steps:
s2.1, based on the improved latent structure model, acquiring statistics of a single sample X in the input data X in a quality related subspace and a quality independent subspace as follows:
where D represents the statistics of the quality-related subspace, T represents the transpose, T m Representing a score vector, Λ representing a second intermediate coefficient;
s2.2, constructing control limits of statistics:
wherein C is D Representing a control limit of a quality-related subspace, F A,n-A,α F distribution with degrees of freedom A and n-A and confidence alpha is represented;
s2.3, judging whether the statistic D of the quality related subspace is larger than the control limit C D If yes, judging that the sample x fails, otherwise, judging that the sample x does not fail;
s2.4, acquiring the fault state of each sample in the input data X in the industrial process according to the method in the steps S2.1-S2.3, and completing the construction flow of the judging standard of the fault state in the industrial process.
The step of establishing a fault library in S3 includes establishing a quality-related fault library and a quality-independent fault library.
The specific steps of establishing the quality related fault library are as follows:
s3.1.1 construction of quality-related fault datasets based on prior experienceWherein k is 1 For the total number of fault types>All represent fault data;
s3.1.2 fault data X f,1 Projection to a quality related subspace and a quality independent subspace, specifically:
wherein T is f,1 A scoring matrix representing the fault data,representing a load matrix->Representing a quality-related fault space resulting from projection of quality-related fault data into a quality-related subspace,/a quality-related fault space resulting from projection of quality-related fault data into a quality-related subspace>Representing a quality-related fault space obtained by projecting quality-related fault data into a quality-related subspace, wherein T represents transposition;
s3.1.3 calculating quality-related fault spaceAnd quality-related subspace->Specifically, the covariance matrix of (a) is:
wherein C is f,1 Representing quality-related fault spacesC represents the quality independent fault space +.>Is a covariance matrix of (a);
s3.1.4 extracting covariance matrix C by generalized principal component analysis f,1 Generalized eigenvector V of sum covariance matrix C f The method comprises the following steps:
V f ＝eig(C f,1 ,C)
wherein eig represents a eigenvalue decomposition operation;
s3.1.5 according to the generalized eigenvector V f Extracting quality-related fault spaceRemoving the normal information in the database to obtain a quality-related fault space after removing the normal information +.>The method comprises the following steps:
s3.1.6 to quality-related fault space after removing normal informationSingular value decomposition is carried out, specifically:
wherein U is f,1 A projection matrix representing a singular value decomposition,right singular value matrix representing singular value decomposition, D f1 Representing the matrix of singular values on the diagonal in descending order;
s3.1.7 screening out projection matrix U f,1 The first L feature vectors make X f,1 Statistics D after sample reconstruction in (a) is smaller than control limit C D Constructing the first L eigenvectors as a failure subspace Σ f,1 ；
S3.1.8 according to the method of steps S3.1.2-S3.1.7, obtainingCorresponding fault subspaceConstructing a quality-related fault library as sigma f ＝(Σ f,1 ，...,Σ f,k )。
The specific steps of establishing the quality irrelevant fault base are as follows:
s3.2.1 based on improved latent structure model, for predictable spaceThe main component is decomposed, specifically:
wherein T is y Representing a predictable spaceScore matrix of->Representing +.>Projection matrix obtained by PCA decomposition +.>The number of principal elements of A y ，A y Rank (Q), rank (x) represents the rank calculation;
s3.2.2 according to the scoring matrix T y Obtaining a load matrix P y And according to the load matrix P y Score matrix T y ObtainingThe method comprises the following steps:
wherein ( -1 The operation of inversion is represented by the expression,representation->Is a complete mass space of (2);
in the present embodiment, the full quality space represents a space after redundant information is removed in the quality-related space.
S3.2.3 according to the load matrix P y Score matrix T y Acquisition ofQuality independent matrix->The method comprises the following steps:
s3.2.4 matrix independent of qualityThe main component is decomposed, specifically:
/>
wherein T is o Representing quality independent matricesScore matrix, P o Representing a load matrix->The number of the medium pivot elements is A-A y ；
In this embodiment, the principal element number, i.e. the component number, is also equal to the rank of the space.
S3.2.5 according to the complete mass spaceQuality independent matrix->Optimizing the improved latent structure model to obtain an optimized latent structure model which is:
s3.2.6 construction of quality independent fault datasets based on prior experienceWherein k is 2 Representing the total number of quality independent fault types, +.>Failure data;
s3.2.7 based on the optimized latent structure model, fault data are obtainedProjection to a quality related subspace and a quality independent subspace, specifically:
wherein,representing quality-related subspace->Is>Representing fault data +.>Corresponding quality-related subspace, < >>Representing fault data +.>Corresponding quality independent matrix, < >>Representation->Score matrix of->Representing fault data/>Corresponding scoring matrix, ">Representing fault data +.>A corresponding quality independent subspace;
s3.2.8 obtaining quality independent matrixAnd quality independent matrix->The covariance matrix of (2) is:
wherein,representing quality independent matrix->Corresponding covariance matrix, C represents quality independent matrix +.>A corresponding covariance matrix;
s3.2.9 extracting covariance matrix by generalized principal component algorithmSum covariance matrixGeneralized eigenvector of CThe method comprises the following steps:
s3.2.10 according to generalized eigenvectorExtracting quality independent matrix->The normal information in (a) is removed, specifically: />
Wherein,representing a fault space after the normal information is removed;
s3.2.11 to fault space after removing normal informationSingular value decomposition is carried out, specifically:
wherein,represents the (u) 1 Total fault subspace of individual quality independent faults, < ->A matrix representing a diagonal of descending singular value arrangement;
s3.2.12 screening out Total Fault subspacesMiddle and top L eigenvectors make +.>Statistics D after sample reconstruction in (a) is smaller than control limit C D The first L eigenvectors are built as failure subspace +.>
S3.2.13 according to the method of steps S3.2.6-S3.2.12, obtainingCorresponding failure subspace->Constructing a quality independent fault library as +.>
The step S4 specifically includes:
s4.1, acquiring a monitoring fault sample x based on a fault database new In quality-related fault bank Σ f The fault amplitude f corresponding to each type of fault i The method comprises the following steps:
wherein i=1, 2,..k 1, ( + Represents the generalized inverse, T represents the transpose, Ω represents the third intermediate coefficient, x new Representing real-time data measured by m sensors in an on-line test and being a failure sample,the representation dimension is m×1;
s4.2, for monitoring fault sample x new Performing fault reconstruction, wherein the reconstructed sample x is specifically as follows:
x * ＝x new -Σ f,i f i
s4.3, traversing the quality-related fault library Σ f Obtaining statistics D of the reconstructed samples and judging whether the statistics D of the reconstructed samples are smaller than or equal to a control limit C D If yes, identifying the fault as an i-th type fault, otherwise, entering a step S4.4;
s4.4, obtaining a monitoring fault sample x new In a quality independent fault bank Σ u The fault amplitude f corresponding to each type of fault i The method comprises the following steps:
s4.5, for monitoring fault sample x new Performing fault reconstruction, and reconstructing a sample x * The method comprises the following steps:
s4.4, traversing the quality independent fault library Σ u Obtaining statistics D of the reconstructed samples and judging whether the statistics D of the reconstructed samples are smaller than or equal to a control limit C D If yes, the fault is identified as false alarm, false alarm feedback is carried out, otherwise, the fault is identified as unknown type fault, and the step S5 is carried out.
The step S5 specifically comprises the following steps:
s5.1, failure sample x is to be monitored new Corresponding output y new The input improved latent structure model is as follows:
wherein,representing a monitoring fault sample x new Quality-related subspace of->Representing a monitoring fault sample x new Quality independent subspace of->Representing the output y new Quality-related subspace of->Representing the output y new Quality independent subspaces of (2);
s5.2, acquiring quality independent subspacesStatistics Q of f,y The method comprises the following steps:
s5.3, obtaining a control limit of each output sample y under normal working conditionsThe method comprises the following steps:
where g 'represents a fourth intermediate coefficient, h' represents a fifth intermediate coefficient,the method is characterized in that chi-square distribution is represented, xi represents the variance of the historical statistics SPE under normal working conditions, and mu represents the mean value of the historical statistics SPE;
s5.4, judging statistic Q f,y Whether or not it is greater than the control limitIf yes, the unrecognizable monitoring fault sample is a quality related fault, otherwise, the unrecognizable monitoring fault sample is a quality unrelated fault.
In this example, the method presented herein was validated by data collected in a tenaci-eastmann (TEP) experiment. TEP is a small industrial process developed by isman chemical company Downs and Vogel in 1993, the whole process consisting of five operating units including chemical reactors, condensers, compressors, vapor/liquid separators and separators.
TEP contains eight components: a, B, C, D, E, F, G and H, wherein gaseous species A, C, D and E and inert species B are reactants, G and H are reaction products, and F is a reaction byproduct. The reaction was carried out simultaneously in the reactor with the aid of a catalyst for 4 reactions:
TABLE 1 15 known faults (IDV)
TEP co-generation into 22 data sets for process monitoring and fault diagnosis, including 1 normal data X and 10 quality-related fault training sets X r And 5 quality independent failure training sets X u . In the training set, the normal data set comprises 480 samples for building a training model and calculating a control limit, and the fault data set comprises 480 fault samples for building a fault library; in the test set, each test data set contains 980 samples, the first 200 being normal samples and the last 780 being faulty samples for experimental verification. Each input sample includes 33 variables, and the test sample is 1 variable.
The fault types are specifically shown in table 1, and the quality related fault and the quality unrelated fault are judged: the SPE statistic of output y is used as a criterion for discrimination. If y fails, then the corresponding output and Q y The control limit is exceeded. For each set of fault samples, assume n y Indicating the number of affected fault samples, n 1 Representing the total number of samples. If n y /n 1 >0.1, the fault is deemed to be y-related. According to this principle, the fault type IDV (1, 2,5-8,10,12,13) is identified as quality related fault data and IDV (3,4,9,11,14,15) is identified as quality independent fault data.
TABLE 2 quality-related fault subspace extraction
/>
Table 2 shows the fault subspaces extracted for 10 classes of quality-related faults in an offline process, from which a quality-related fault library Σ is composed r . The fault subspaces contain the fault characteristics, and fault information can be completely removed in the reconstruction process, so that a fault sample returns to a normal value. FIG. 2 shows a specific process for reconstruction of the fault 8 in different dimensions, where the line is T 2 Statistics, dashed lines are control limits. It can be seen that the statistics of the fault samples do not change significantly when the fault subspace is only one-dimensional. When a two-dimensional fault direction is used, the statistics are obviously reduced, and when a three-dimensional fault direction is used, the statistics of fault samples are all reduced to be below a control limit, namely the normal value is returned.
The fault library of the offline process is used for the identification and diagnosis of online faults. First, fault 1,6,7 is reconstructed by using the fault subspace of fault 2 in the fault library, and the result is shown in FIG. 3, wherein the solid line is T 2 Statistics, point reconstructed T 2 Statistics, dashed lines are control limits, and obviously, each fault cannot be reconstructed by fault 2.
The corresponding faults are reconstructed by adopting the fault subspaces of the faults 1,2 and 13 respectively, and the result is shown in fig. 4. Obviously, the fault type matching the fault subspace can be fully reconstructed.
For quality independent faults, verifying the reconstruction performance of the quality independent faults by adopting a quality independent fault library established by an offline process; the results are shown in FIG. 5.
As shown in fig. 5, the first 5 diagrams are monitoring cases of quality independent faults, and it can be seen that there are partial false alarms in the monitoring of quality independent faults. When such a situation occurs, it cannot be reconstructed by a quality-related fault, but is reconstructed with a quality-independent fault, the result being shown in the last 5 figures. The quality irrelevant fault library can be seen to reconstruct the corresponding quality irrelevant faults effectively, and further, false alarms are identified and timely feedback is carried out.
Claims (7)
1. The quality-related fault diagnosis and false alarm feedback method based on fault reconstruction is characterized by comprising the following steps of:
s1, establishing an improved latent structure model;
s2, constructing a judging standard of a fault state in an industrial process based on the improved latent structure model;
s3, establishing a fault library according to the judging standard of the fault state;
s4, based on a fault library, carrying out fault identification and false alarm feedback on the monitored fault sample;
s5, extracting a fault subspace of the unrecognizable monitoring fault sample, and updating a fault library according to the fault subspace;
the step S1 specifically comprises the following steps:
s1.1, collecting a plurality of input data X and output data Y in an industrial process, wherein a column of the input data X comprises m variables, and a row of the input data X comprises n sampling samples; the column of output data Y includes p variables, the row of which includes n sample samples;
s1.2, setting the iteration number i=1;
s1.3, randomly fetching a column of data u of the output data Y i Let the first intermediate coefficient u old ＝u i Data u i The updating is carried out specifically as follows:
|w i |＝w i /norm(wi)
t i ＝X i |w i |/|w i | T |w i |
q i ＝Y i t i /t i T t i
wherein u is i ' represents updated data u i ，Y i Representing the output Y, q after the ith update i Representing coefficient vectors, T representing transpose, X i Representing X after the ith update i ，t i Representing data X i Along the modulus |w i Dimension reduction data after projection, |w i The I represents the regression vector w i Norm (x) represents the operation of taking the binary norm;
s1.4, judging the I U old -u i '||/||u old ||<10 -4 If yes, entering a step S1.5, otherwise, adding one to the count value of i, and returning to the step S1.3;
s1.5 according to data X i Dimension t decreases i Obtaining a load vector p i The method comprises the following steps:
s1.6 according to the load vector p i Coefficient directionQuantity q i Dimension t decreases i Data X i Data Y i The updating is carried out specifically as follows:
wherein X is i+1 Representing updated data X i ，Y i+1 Representing updated Y i '；
S1.7, judging whether the iteration times i are equal to A, if so, entering a step S1.8, otherwise, adding one to the count value of i, and returning to the step S1.3;
s1.8, the A-time updated reduced data is formed into a scoring matrix T= (T) 1 ,...,t A ) The regression vector updated a times is formed into a weight matrix w= (W) 1 ,...,w A ) The load vector updated a times is formed into a load matrix p= (P) 1 ,...,p A )；
S1.9, according to the scoring matrix T, the weight matrix W and the load matrix P, establishing an improved latent structure model as follows:
wherein,representing quality-related subspaces, +.>Representing quality independent subspaces, T m Represented by X along P m Score matrix obtained by dimension reduction, P m Representing projection vectors, M representing a first coefficient matrix; />Representing a predictable space, which is information directly affected by X; />Representing unpredictable space, which is space not directly affected by X; r represents a second coefficient matrix, Q represents a load matrix of Y,>representing residual projection vector, Λ m Representing a matrix of eigenvalues.
2. The quality-related fault diagnosis and false alarm feedback method based on fault reconstruction according to claim 1, wherein the step S2 specifically comprises:
s2.1, based on the improved latent structure model, acquiring statistics of a single sample X in the input data X in a quality related subspace and a quality independent subspace as follows:
where D represents the statistics of the quality-related subspace, T represents the transpose, T m Representing a score vector, Λ representing a second intermediate coefficient;
s2.2, constructing control limits of statistics:
wherein C is D Representing a control limit of a quality-related subspace, F A,n-A,α F distribution with degrees of freedom A and n-A and confidence alpha is represented;
s2.3, judging whether the statistic D of the quality related subspace is larger than the control limit C D If yes, judging that the sample x fails, otherwise, judging that the sample x does not fail;
s2.4, acquiring the fault state of each sample in the input data X in the industrial process according to the method in the steps S2.1-S2.3, and completing the construction flow of the judging standard of the fault state in the industrial process.
3. The quality-related fault diagnosis and false alarm feedback method based on fault reconstruction according to claim 2, wherein the step S3 of establishing a fault library includes establishing a quality-related fault library and a quality-independent fault library.
4. The quality-related fault diagnosis and false alarm feedback method based on fault reconstruction of claim 3, wherein the specific steps of establishing the quality-related fault library are as follows:
s3.1.1 construction of quality-related fault datasets based on prior experienceWherein k is 1 For the total number of fault types>All represent fault data;
s3.1.2 fault data X f,1 Projection to a quality related subspace and a quality independent subspace, specifically:
wherein T is f,1 A scoring matrix representing the fault data,representing a load matrix->Representing a quality-related fault space resulting from projection of quality-related fault data into a quality-related subspace,/a quality-related fault space resulting from projection of quality-related fault data into a quality-related subspace>Representing a quality-related fault space obtained by projecting quality-related fault data into a quality-related subspace, wherein T represents transposition;
s3.1.3 calculating quality-related fault spaceAnd quality-related subspace->Specifically, the covariance matrix of (a) is:
wherein C is f,1 Representing quality-related fault spacesC represents the quality independent fault space +.>Is a covariance matrix of (a);
s3.1.4 extracting covariance matrix C by generalized principal component analysis f,1 Generalized eigenvector V of sum covariance matrix C f The method comprises the following steps:
V f ＝eig(C f,1 ,C)
wherein eig represents a eigenvalue decomposition operation;
s3.1.5 according to the generalized eigenvector V f Extracting quality-related fault spaceRemoving the normal information in the database to obtain a quality-related fault space after removing the normal information +.>The method comprises the following steps:
s3.1.6 to quality-related fault space after removing normal informationSingular value decomposition is carried out, specifically:
wherein U is f,1 A projection matrix representing a singular value decomposition,right singular value matrix representing singular value decomposition, D f,1 Representing the matrix of singular values on the diagonal in descending order;
s3.1.7 screening out projection matrix U f,1 The first L feature vectors make X f,1 Statistics D after sample reconstruction in (a) is smaller than control limit C D Will beThe first L eigenvectors are built into the failure subspace Σ f,1 ；
S3.1.8 according to the method of steps S3.1.2-S3.1.7, obtainingCorresponding fault subspaceConstructing a quality-related fault library as sigma f ＝(Σ f,1 ，…,Σ f,k )。
5. The quality-related fault diagnosis and false alarm feedback method based on fault reconstruction according to claim 4, wherein the specific steps of establishing a quality-independent fault library are as follows:
s3.2.1 based on improved latent structure model, for predictable spaceThe main component is decomposed, specifically:
wherein T is y Representing a predictable spaceScore matrix of->Representing +.>Projection matrix obtained by PCA decomposition +.>Is the main part of (2)The number of the elements is A y ，A y Rank (Q), rank (x) represents the rank calculation;
s3.2.2 according to the scoring matrix T y Obtaining a load matrix P y And according to the load matrix P y Score matrix T y ObtainingThe method comprises the following steps:
wherein ( -1 The operation of inversion is represented by the expression,representation->Is a complete mass space of (2);
s3.2.3 according to the load matrix P y Score matrix T y Acquisition ofQuality independent matrix->The method comprises the following steps:
s3.2.4 matrix independent of qualityThe main component is decomposed, specifically:
wherein T is o Representing quality independent matricesScore matrix, P o Representing a load matrix->The number of the medium pivot elements is A-A y ；
S3.2.5 according to the complete mass spaceQuality independent matrix->Optimizing the improved latent structure model to obtain an optimized latent structure model which is:
s3.2.6 construction of quality independent fault datasets based on prior experienceWherein k is 2 Representing the total number of quality independent fault types, +.>Failure data;
s3.2.7 based on the optimized latent structure model, fault data are obtainedProjection to a quality related subspace and a quality independent subspace, specifically:
wherein,representing quality-related subspace->Is>Representing fault data X u1 Corresponding quality-related subspace, < >>Representing fault data +.>Corresponding quality independent matrix, < >>Representation->Score matrix of->Representing fault data +.>Corresponding scoring matrix, ">Representing fault data +.>A corresponding quality independent subspace;
s3.2.8 obtaining quality independent matrixAnd quality independent matrix->The covariance matrix of (2) is:
wherein,representing quality independent matrix->Corresponding covariance matrix, C represents quality independent matrix +.>A corresponding covariance matrix;
s3.2.9 extracting covariance matrix by generalized principal component algorithmGeneralized eigenvector +.>The method comprises the following steps:
s3.2.10 according to generalized eigenvectorExtracting quality independent matrix->The normal information in (a) is removed, specifically:
wherein,representing a fault space after the normal information is removed;
s3.2.11 to fault space after removing normal informationSingular value decomposition is carried out, specifically:
wherein,represents the (u) 1 Total fault subspace of individual quality independent faults, < ->A matrix representing a diagonal of descending singular value arrangement;
s3.2.12 screening out Total Fault subspacesMiddle and top L eigenvectors make +.>Statistics D after sample reconstruction in (a) is smaller than control limit C D The first L eigenvectors are built as failure subspace +.>
S3.2.13 according to the method of steps S3.2.6-S3.2.12, obtainingCorresponding fault subspaceConstructing a quality independent fault library as +.>
6. The quality-related fault diagnosis and false alarm feedback method based on fault reconstruction according to claim 5, wherein the step S4 specifically comprises:
s4.1, acquiring a monitoring fault sample x based on a fault database new In quality-related fault bank Σ f The fault amplitude f corresponding to each type of fault i The method comprises the following steps:
wherein i=1, 2,..k 1, ( + Represents the generalized inverse, T represents the transpose, Ω represents the third intermediate coefficient, x new Representing real-time data measured by m sensors in an on-line test and being a failure sample, the representation dimension is m×1;
s4.2, for monitoring fault sample x new Performing fault reconstruction, and reconstructing a sample x * The method comprises the following steps:
x * ＝x new -Σ f,i f i
s4.3, traversing the quality-related fault library Σ f Obtaining statistics D of the reconstructed samples and judging whether the statistics D of the reconstructed samples are smaller than or equal to a control limit C D If yes, identifying the fault as an i-th type fault, otherwise, entering a step S4.4;
s4.4, obtaining a monitoring fault sample x new In a quality independent fault bank Σ u The fault amplitude f corresponding to each type of fault i The method comprises the following steps:
s4.5, for monitoring fault sample x new Performing fault reconstruction, and reconstructing a sample x * The method comprises the following steps:
s4.4, traversing the quality independent fault library Σ u Obtaining statistics D of the reconstructed samples and judging whether the statistics D of the reconstructed samples are smaller than or equal to a control limit C D If yes, the fault is identified as false alarm, false alarm feedback is carried out, otherwise, the fault is identified as unknown type fault, and the step S5 is carried out.
7. The quality-related fault diagnosis and false alarm feedback method based on fault reconstruction according to claim 6, wherein the step S5 specifically comprises:
s5.1, failure sample x is to be monitored new Corresponding output y new The input improved latent structure model is as follows:
wherein,representing a monitoring fault sample x new Quality-related subspace of->Representing a monitoring fault sample x new Quality independent subspace of->Representing the output y new Quality-related subspace of->Representing the output y new Quality independent subspaces of (2);
s5.2, acquiring quality independent subspacesStatistics Q of f,y The method comprises the following steps:
s5.3, obtaining a control limit of each output sample y under normal working conditionsThe method comprises the following steps:
where g 'represents a fourth intermediate coefficient, h' represents a fifth intermediate coefficient,the method is characterized in that chi-square distribution is represented, xi represents the variance of the historical statistics SPE under normal working conditions, and mu represents the mean value of the historical statistics SPE;
s5.4, judging statistic Q f,y Whether or not it is greater than the control limitIf yes, the unrecognizable monitoring fault sample is a quality related fault, otherwise, the unrecognizable monitoring fault sample is a quality unrelated fault.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110333398.0A CN113076211B (en) | 2021-03-29 | 2021-03-29 | Quality-related fault diagnosis and false alarm feedback method based on fault reconstruction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110333398.0A CN113076211B (en) | 2021-03-29 | 2021-03-29 | Quality-related fault diagnosis and false alarm feedback method based on fault reconstruction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113076211A CN113076211A (en) | 2021-07-06 |
| CN113076211B true CN113076211B (en) | 2024-02-23 |
Family
ID=76611066
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110333398.0A Active CN113076211B (en) | 2021-03-29 | 2021-03-29 | Quality-related fault diagnosis and false alarm feedback method based on fault reconstruction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113076211B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5465321A (en) * | 1993-04-07 | 1995-11-07 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Hidden markov models for fault detection in dynamic systems |
| CN110083860A (en) * | 2019-03-13 | 2019-08-02 | 东北大学 | A kind of industrial method for diagnosing faults based on correlated variables selection |
| CN110221590A (en) * | 2019-05-17 | 2019-09-10 | 华中科技大学 | A kind of industrial process Multiple faults diagnosis approach based on discriminant analysis |
| CN110929765A (en) * | 2019-11-06 | 2020-03-27 | 北京工业大学 | Convolution self-coding fault monitoring method based on batch imaging |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9053291B2 (en) * | 2010-09-29 | 2015-06-09 | Northeastern University | Continuous annealing process fault detection method based on recursive kernel principal component analysis |
-
2021
- 2021-03-29 CN CN202110333398.0A patent/CN113076211B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5465321A (en) * | 1993-04-07 | 1995-11-07 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Hidden markov models for fault detection in dynamic systems |
| CN110083860A (en) * | 2019-03-13 | 2019-08-02 | 东北大学 | A kind of industrial method for diagnosing faults based on correlated variables selection |
| CN110221590A (en) * | 2019-05-17 | 2019-09-10 | 华中科技大学 | A kind of industrial process Multiple faults diagnosis approach based on discriminant analysis |
| CN110929765A (en) * | 2019-11-06 | 2020-03-27 | 北京工业大学 | Convolution self-coding fault monitoring method based on batch imaging |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113076211A (en) | 2021-07-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jang et al. | Adversarial autoencoder based feature learning for fault detection in industrial processes | |
| CN108062565B (en) | Double-principal element-dynamic core principal element analysis fault diagnosis method based on chemical engineering TE process | |
| Zhao et al. | Fault-relevant principal component analysis (FPCA) method for multivariate statistical modeling and process monitoring | |
| Wang et al. | Data-driven soft sensor approach for quality prediction in a refining process | |
| CN110738274A (en) | nuclear power device fault diagnosis method based on data driving | |
| CN113642754B (en) | Complex industrial process fault prediction method based on RF noise reduction self-coding information reconstruction and time convolution network | |
| CN108549908B (en) | Chemical process fault detection method based on multi-sampling probability kernel principal component model | |
| CN111340110B (en) | Fault early warning method based on industrial process running state trend analysis | |
| Gao et al. | A process fault diagnosis method using multi‐time scale dynamic feature extraction based on convolutional neural network | |
| CN112000081B (en) | Fault monitoring method and system based on multi-block information extraction and Mahalanobis distance | |
| Luo et al. | Key-performance-indicators-related fault subspace extraction for the reconstruction-based fault diagnosis | |
| CN111639304B (en) | CSTR fault positioning method based on Xgboost regression model | |
| Huang et al. | Double‐step block division plant‐wide fault detection and diagnosis based on variable distributions and relevant features | |
| CN113703422B (en) | Gas turbine pneumatic actuator fault diagnosis method based on feature analysis processing | |
| Ruiz et al. | Multivariate Principal Component Analysis and Case-Based Reasoning for monitoring, fault detection and diagnosis in a WWTP | |
| CN114297921A (en) | AM-TCN-based fault diagnosis method | |
| Wang et al. | Fault monitoring based on adaptive partition non-negative matrix factorization for non-Gaussian processes | |
| Li et al. | Multiple sensor fault diagnosis for dynamic processes | |
| CN113076211B (en) | Quality-related fault diagnosis and false alarm feedback method based on fault reconstruction | |
| Dong et al. | Dynamic-inner canonical correlation analysis based process monitoring | |
| CN110794814B (en) | Fault determination method and system based on generalized principal component | |
| CN114638039A (en) | Structural health monitoring characteristic data interpretation method based on low-rank matrix recovery | |
| Shang et al. | Sparse slow feature analysis for enhanced control monitoring and fault isolation | |
| Guo et al. | Fault Diagnosis Combining Information Entropy with Transfer Entropy for Chemical Processes | |
| CN114112374B (en) | Fault diagnosis method of weapon equipment system based on orthogonal sub-block projection |
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 |