CN110297475B - Intermittent process fault monitoring method based on fourth-order moment singular value decomposition - Google Patents
Intermittent process fault monitoring method based on fourth-order moment singular value decomposition Download PDFInfo
- Publication number
- CN110297475B CN110297475B CN201910664867.XA CN201910664867A CN110297475B CN 110297475 B CN110297475 B CN 110297475B CN 201910664867 A CN201910664867 A CN 201910664867A CN 110297475 B CN110297475 B CN 110297475B
- Authority
- CN
- China
- Prior art keywords
- data
- moment
- line
- follows
- formula
- 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]
- G05B19/41875—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 quality surveillance of production
-
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- 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 an intermittent process fault monitoring method based on fourth-order moment singular value decomposition, which is used for solving the problem of data nonlinearity in an intermittent process and non-Gaussian property caused by nonlinearity. The invention comprises two stages of an off-line modeling stage and an on-line monitoring stage. The "offline modeling phase" includes: firstly, standardizing data, performing fourth-order moment processing, and combining a fourth-order moment matrix; and then singular value decomposition is carried out, and the obtained matrix is simplified to prepare for monitoring. The "on-line monitoring phase" includes: standardizing the online data, performing fourth-order moment processing, and combining a fourth-order moment matrix; then calculating the statistic and the residual error and the corresponding control line; and finally, monitoring the generation process by using the statistic, and generating an alarm when a fault is found. The invention fully considers the nonlinearity and non-Gaussian property of the intermittent process data, reduces the false alarm rate in the normal stage, reduces the false alarm rate in the fault stage, accelerates the response speed and has higher practical value.
Description
Technical Field
The invention belongs to the field of industrial process fault monitoring, and particularly relates to a four-order moment singular value decomposition technology. The method for monitoring the fault based on the fourth-order moment singular value decomposition is specifically applied to the TE (Tennessee Eastman) process.
Background
Modern industrial processes have a large number of intermittent processes, and common intermittent processes include microbial pharmacy, sewage treatment, beer preparation, yoghourt preparation and the like. Batch production in the intermittent process is flexible in scale, the process is easy to change, meanwhile, the product switching has certain compatibility, a small amount of production of different varieties can be carried out, and the method can adapt to the change of raw materials or operation conditions quickly.
Due to the nonlinearity of the system, the collected data generally has a non-Gaussian distribution, and non-Gaussian information is very important for monitoring the system. Typically, non-gaussian information requires high order analysis (data order greater than 2).
Currently, the high-order analysis methods mainly include: kernel Principal Component Analysis (KPCA), Multivariate Kernel Independent Component Analysis (MKICA), Multivariate Kernel Entropy Component Analysis (MKECA). The high-order analysis method used by the above algorithm is a kernel technique. The kernel can map data to high dimension, but structural information among the data is damaged at the same time, and fault diagnosis is influenced to a certain extent.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides an intermittent process fault monitoring method based on fourth-order moment singular value decomposition. Each variable generates a fourth moment with itself at the previous time, rather than operating with the entire data through a kernel. Thereby preserving the structure of the underlying data itself. The fourth-order moment contains remarkable non-Gaussian information, and the monitoring accuracy is greatly improved. The method presented herein performs fourth-order moment processing on the data prior to performing the statistics-based monitoring. In the stage of constructing the statistic on the data, the statistic naturally has high-order characteristics due to the fact that the data have high-order characteristics, and a new statistic mode does not need to be additionally created.
The invention adopts the following technical scheme and implementation steps:
A. an off-line modeling stage:
1) reading in normal data, calculating mean value mean of each (total D) variabledAnd standard deviation stddNormalized to normal data, the formula is as follows:
wherein the content of the first and second substances,data representing all the time instants of the d-th variable,to representIs determined by the average value of (a) of (b),to representStandard deviation of (d);
2) for a total of D speciesThe data of each moment is processed by fourth moment, and the formula is as follows;
representing the fourth moment at the kth instant of the d-th variable, k representing the sampling instant.Indicating the value of the d-th variable at the k-th time.Denotes the step size, 1, 2, 3 are selected in this text, under which conditions。
wherein N represents an end time;
4) singular Value Decomposition (SVD) of C, SVD (C) = USVTA two-step simplification of U is performed.
i.U, the first step is simplified as follows:
the minimum value of M that the following formula can satisfy is calculated.
Wherein the content of the first and second substances,is the element on the diagonal of S, and I is the minimum in the number of S rows and columns.Is a threshold value, which can be adjusted, here 90.
And (5) reserving the first M columns of the U, and deleting the rest to obtain the U after the first step of simplification.
A second step simplification of u, steps as follows:
judging the numerical value of each element in the simplified U, wherein the formula is as follows:
wherein the content of the first and second substances,representing the square of the ith row and mth column element in U,representing the sum of the squares of all elements in row i.The number of columns of U before deletion.
When in useAnd (4) satisfying the judgment condition of the formula above, and setting the judgment condition to be 0 to obtain the simplified U in the second step.
5) And cutting the S matrix into M rows and M columns. Inverse matrix S of the clipped S is solvedinv. U and after two steps in the foregoing are simplifiedSinvPreservation for later on-line monitoring
B. And (3) an online monitoring stage:
6) reading in on-line data and standardizing the on-line data, wherein the formula is as follows:
wherein the content of the first and second substances,for the value of the d-th variable of the online data,mean and standard deviation of normal data;
and combined into Con
wherein the content of the first and second substances,the confidence coefficient is 95%, the numerator degree of freedom M (the value in step 4), and the denominator degree of freedom-F-test of M.。
Wherein the content of the first and second substances,the matrix is an S matrix obtained after the on-line data is subjected to singular value decomposition. L is the value of the element on its diagonal.Is in LThe ith element. c is the 95% quantile expected to be 0 with a standard deviation of 1.
10) At each momentIn contrast to the corresponding control line. If not, returning to the step 6; if the control line is exceeded, a fault occurs, and an alarm is given.
Advantageous effects
Compared with the prior art, the method for constructing the high-order statistics keeps the information of the original structure of the data, and fully excavates the information of the high-order data in the monitoring process. The nonlinearity of the data and the non-Gaussian property caused by the nonlinearity are fully considered. Data are analyzed in a mode of constructing the data into fourth moment, and statistics and residual errors of the statistics are calculated by the analyzed data, so that a satisfactory monitoring effect is achieved.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2(a) shows TE process fault 1F2Monitoring a graph;
FIG. 2(b) is a TE process fault 1 FS monitoring diagram;
FIG. 3(a) is a TE process fault 2F2Monitoring a graph;
FIG. 3(b) is a TE process fault 2 FS monitoring diagram;
FIG. 4(a) shows TE process fault 5F2Monitoring a graph;
FIG. 4(b) is a TE process fault 5 FS monitor diagram;
FIG. 5(a) is TE process failure 1 MKICA I2A statistic monitoring graph;
FIG. 5(b) is a TE process fault 1 MKICA residual monitoring diagram;
Detailed Description
The algorithm presented herein can monitor faults that occur in industrial batch process production. And performing singular value decomposition by processing the variable fourth moment, and constructing statistics and corresponding control lines to complete monitoring. And finally, a satisfactory monitoring result can be obtained, and the safe production is ensured.
To verify the accuracy of the algorithms presented herein, tests were performed using TE process data. The TE (Tennessee Eastman) simulation platform is a simulation platform established according to the actual chemical reaction process, the generated data has time-varying, strong coupling and nonlinear characteristics, and the TE (Tennessee Eastman) simulation platform is widely used for testing control and fault diagnosis models of complex industrial processes.
There are 21 kinds of failures in TE process, and the detailed description is shown in Table 1
TABLE 1 TE Process Fault List
Fault numbering | Cause of failure |
1 | A, C feed rate, B composition constant (stream 4) |
2 | Composition B, constant feed ratio of A, C (stream 4) |
3 | D feed temperature (stream 2) |
4 | Reactor cooling water entry temperature |
5 | Condenser cooling water entry temperature |
6 | A feed reduction (stream 1) |
7 | C head pressure reduction (stream 4) |
8 | A, B, C feed combination (stream 4) |
9 | D feed temperature (stream 2) |
10 | C feed temperature (stream 4) |
11 | Reactor cooling |
12 | Condenser cooling water entry temperature |
13 | Kinetics of the reaction |
14 | Cooling water valve of |
15 | Cooling water valve of condenser |
16 | Is unknown |
17 | Is unknown |
18 | Is unknown |
19 | Is unknown |
20 | Is unknown |
21 | The valve of flow 4 is fixed |
The method comprises the following specific implementation steps:
A. an off-line modeling stage:
step 1: reading in normal data, calculating mean value mean of each (total D) variabledAnd standard deviation stddNormalized to normal data, the formula is as follows:
Step 2: by the formulaTo XdThe data at each moment in time of (2) is subjected to fourth moment processing. k is time, startHas a value of;
And step 3: will be provided withThe combination is a fourth moment matrix C, and the spelling method is as follows:
And 4, step 4: and svd decomposition is carried out on the C to obtain a U matrix and an S matrix, and two-step simplification is carried out.
The first step is as follows: find the equation of the commandMinimum M value satisfied. The first M columns of U are retained, and the remainder are deleted.
The second step is that: according to the formulaJudging each element in U. The value of i starts from 1 and is the maximum number of rows of U. The value of M is from 1 and is at most M. Will satisfy the conditionsAnd setting 0.
And 5: and cutting the S matrix into M rows and M columns, and deleting the rest. Inverse matrix S of the clipped S is solvedinv;
B. And (3) an online monitoring stage:
step 6: according to the formulaThe on-line data is standardized and,mean and standard deviation of normal data. D ranges from 1 to D. (ii) a
And 7: according to the formulaAnd performing fourth-order moment processing on the online data. And are combined into
k is the current time and the initial value isTherefore, it is necessary to go to the second step in the industrial processStarting monitoring after each moment;
the statistic calculation formula at all the moments is as follows:
the statistic calculation formula for the k-th time alone is as follows:
wherein the content of the first and second substances,is composed ofThe k-th column of (1).The statistic and residual error at time k are shown.
And step 9: the control line for calculating the statistic and the residual error is as follows:
step 10: at each momentIn contrast to the corresponding control line. If not, returning to the step 6; if the control line is exceeded, a fault occurs, and an alarm is given.
The steps are the specific application steps of the method in the TE process.
The validity of the method can be verified by means of fig. 2, 3 and 4.
Before time 200, production was in the normal phase. It can be seen that the statistics and residual values proposed by the method herein are all lower than the control line, and no false positives are generated. As can be seen from fig. 5, MKICAs have more false alarms in the early stage due to the large fluctuation in the initial stage of the industrial process.
The data of the TE process introduced a fault at time 200. As can be seen from fig. 2, 3 and 4, since the statistics constructed herein are of fourth order nature, and therefore very sensitive to fault response, the control line is crossed over a short period of time, generating an alarm. The production safety is successfully ensured. The high-order analysis mode used by MKICA is a nuclear technique, and no four-order moment method is effective for mining high-order information of data. Therefore, it can be seen from the graph that the MKICA statistic rises more slowly, without the high-speed response capability of the fourth-order moment statistic. In an actual industrial process, timely response is a very important capability.
The analysis of the experimental graph can be used for obtaining that the intermittent process fault diagnosis method based on the fourth-order moment singular value decomposition has a lower false alarm rate in a normal stage and a lower false negative rate and a faster response capability in a fault stage. The effectiveness of the proposed method is demonstrated.
Claims (1)
1. A method for monitoring intermittent process faults based on fourth-order moment singular value decomposition is characterized by comprising an off-line modeling stage and an on-line monitoring stage, and comprises the following specific steps:
A. an off-line modeling stage:
1) reading in normal data, calculating mean value mean of each variabledAnd standard deviation stddNormalized to normal data, the formula is as follows:
wherein, XdData representing all the times of the d-th variable, meandRepresents XdAverage value of (std)dRepresents XdThe standard deviation of (a), there are D variables;
2) for X which is totally D and standardizeddThe data of each moment is processed by fourth moment, and the formula is as follows;
cd(k)=xd(k)xd(k-τ1)xd(k-τ2)xd(k-τ3)
cd(k) representing the fourth moment of the kth moment of the d-th variable, k representing the sampling moment, xd(k) Denotes the value of the d-th variable at the k-th instant, τ1、τ2、τ3Represents a step size;
3) c is tod(k) The combination is a fourth moment matrix C, and the formula is as follows:
wherein N represents an end time;
4) singular Value Decomposition (SVD), SVD (C) USV for CTThe U is simplified in two steps,
i.U, the first step is simplified as follows:
the minimum value of M that can be satisfied by the following formula is calculated,
wherein S isi,iIs the element on the diagonal of S, I is the minimum value in the S row and column number, and delta is the threshold value which can be adjusted;
reserving the first M columns of the U, and deleting the rest to obtain the U after the first step of simplification;
a second step simplification of u, steps as follows:
judging the numerical value of each element in the simplified U, wherein the formula is as follows:
wherein the content of the first and second substances,representing the square of the element in row i and column m in the simplified U,representing the sum of squares of all elements in the ith row in the simplified U, wherein M' is the number of columns of the U before deletion;
when u isi,mThe judgment condition of the formula above is met, and the judgment condition is set to 0 to obtain the simplified U in the second step;
5) cutting the S matrix into M rows and M columns, and obtaining the inverse matrix S of the S after cuttinginvU and S simplified by the two steps in the foregoinginvStoring for later on-line monitoring;
B. and (3) an online monitoring stage:
6) reading in on-line data and standardizing the on-line data, wherein the formula is as follows:
wherein x isd.onFor the value of the d-th variable of the online data, meand,stddMean and standard deviation of normal data;
7) for normalized xd,onFourth moment processing is carried out, and the formula is as follows:
cd,on(k)=xd,on(k)xd,on(k-τ1)xd,on(k-τ2)xd,on(k-τ3)
and combined into Con
8) Computing statistic F2And residual error FS:
F2=(ConU)Sinv(ConU)T
FS=(Con-ConU)(Con-ConU)T
u, S thereininvFor the matrix saved in step 5 of the off-line phase,
where F (0.95, M, N ' -M) represents a confidence of 95%, and the numerator degree of freedom M, which is the value in step 4, is the F test of the denominator degree of freedom N ' -M, N ' ═ N- τ3,
Wherein the content of the first and second substances,
L=diag(Son)
wherein S isonIs an S matrix obtained by performing singular value decomposition on the online data, L is an element value on a diagonal line thereof, LiIs the ith element in L, c is the desired 95% quantile with a standard deviation of 1, 0;
10) f at each moment2FS is compared with the corresponding control line, if not, the control line is normal, and the step 6 is returned; if the control line is exceeded, a fault occurs, and an alarm is given.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910664867.XA CN110297475B (en) | 2019-07-23 | 2019-07-23 | Intermittent process fault monitoring method based on fourth-order moment singular value decomposition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910664867.XA CN110297475B (en) | 2019-07-23 | 2019-07-23 | Intermittent process fault monitoring method based on fourth-order moment singular value decomposition |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110297475A CN110297475A (en) | 2019-10-01 |
CN110297475B true CN110297475B (en) | 2021-07-02 |
Family
ID=68031644
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910664867.XA Active CN110297475B (en) | 2019-07-23 | 2019-07-23 | Intermittent process fault monitoring method based on fourth-order moment singular value decomposition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110297475B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111079857B (en) * | 2019-12-30 | 2023-06-02 | 北京工业大学 | Sewage treatment process fault monitoring method based on overcomplete width learning model |
CN114496209B (en) * | 2022-02-18 | 2022-09-27 | 青岛市中心血站 | Intelligent decision-making method and system for blood donation |
CN115372294A (en) * | 2022-09-15 | 2022-11-22 | 中国市政工程东北设计研究总院有限公司 | Graphite tube stability discrimination method |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20020028626A (en) * | 2000-10-11 | 2002-04-17 | 윤종용 | Fault diagnosis apparatus of nonlinear system based on dynamic threshold and method thereof |
CN103853152A (en) * | 2014-03-21 | 2014-06-11 | 北京工业大学 | Batch process failure monitoring method based on AR-PCA (Autoregressive Principal Component Analysis) |
WO2017164946A1 (en) * | 2016-03-23 | 2017-09-28 | Nec Laboratories America, Inc. | Early warning prediction system |
CN107895224A (en) * | 2017-10-30 | 2018-04-10 | 北京工业大学 | A kind of MKECA fermentation process fault monitoring methods based on extension nuclear entropy load matrix |
CN109828552A (en) * | 2019-02-22 | 2019-05-31 | 北京工业大学 | A kind of batch process Fault monitoring and diagnosis method based on width learning system |
CN110032799A (en) * | 2019-04-15 | 2019-07-19 | 北京工业大学 | A kind of the angle similarity divided stages and monitoring method of microbiological pharmacy process |
-
2019
- 2019-07-23 CN CN201910664867.XA patent/CN110297475B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20020028626A (en) * | 2000-10-11 | 2002-04-17 | 윤종용 | Fault diagnosis apparatus of nonlinear system based on dynamic threshold and method thereof |
CN103853152A (en) * | 2014-03-21 | 2014-06-11 | 北京工业大学 | Batch process failure monitoring method based on AR-PCA (Autoregressive Principal Component Analysis) |
WO2017164946A1 (en) * | 2016-03-23 | 2017-09-28 | Nec Laboratories America, Inc. | Early warning prediction system |
CN107895224A (en) * | 2017-10-30 | 2018-04-10 | 北京工业大学 | A kind of MKECA fermentation process fault monitoring methods based on extension nuclear entropy load matrix |
CN109828552A (en) * | 2019-02-22 | 2019-05-31 | 北京工业大学 | A kind of batch process Fault monitoring and diagnosis method based on width learning system |
CN110032799A (en) * | 2019-04-15 | 2019-07-19 | 北京工业大学 | A kind of the angle similarity divided stages and monitoring method of microbiological pharmacy process |
Also Published As
Publication number | Publication date |
---|---|
CN110297475A (en) | 2019-10-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110297475B (en) | Intermittent process fault monitoring method based on fourth-order moment singular value decomposition | |
CN108227653B (en) | Large-scale nonlinear process monitoring method based on randomized kernel principal component analysis | |
CN108062565B (en) | Double-principal element-dynamic core principal element analysis fault diagnosis method based on chemical engineering TE process | |
CN101689051B (en) | abnormal factor specifying method and system | |
CN107895224B (en) | MKECA fermentation process fault monitoring method based on extended nuclear entropy load matrix | |
CN103853152A (en) | Batch process failure monitoring method based on AR-PCA (Autoregressive Principal Component Analysis) | |
CN101403923A (en) | Course monitoring method based on non-gauss component extraction and support vector description | |
CN111324110A (en) | Fermentation process fault monitoring method based on multiple shrinkage automatic encoders | |
CN112904810B (en) | Process industry nonlinear process monitoring method based on effective feature selection | |
WO2021114320A1 (en) | Wastewater treatment process fault monitoring method using oica-rnn fusion model | |
CN110119579B (en) | OICA-based complex industrial process fault monitoring method | |
CN113657454A (en) | Autoregressive BiGRU-based nuclear power rotating machine state monitoring method | |
CN116627116B (en) | Process industry fault positioning method and system and electronic equipment | |
CN111188761B (en) | Monitoring method for pump equipment based on Fourier-CVA model | |
CN112214006A (en) | Intermittent process fault detection method and system considering two-dimensional dynamic characteristics | |
CN111474920A (en) | ICA and ACVA-based fault detection method | |
CN113253682B (en) | Nonlinear chemical process fault detection method | |
CN111695581B (en) | Nonlinear dynamic process monitoring method based on autoregressive extreme learning machine | |
CN111027645B (en) | Microbial fermentation process fault monitoring method based on high-order statistic construction in sub-stage | |
CN115470838A (en) | Robust dynamic variational Bayesian dictionary learning-based process monitoring method | |
CN109947082B (en) | Process monitoring method based on integrated core independent component analysis model | |
KR20210013840A (en) | Process monitoring apparatus and method | |
Ma et al. | Process monitoring of the pneumatic control valve using canonical variate analysis | |
CN113341721B (en) | Data-driven robust fault diagnosis method facing industrial system containing unknown disturbance | |
CN112270363A (en) | Performance performance index determining method for local neighbor online adaptive modeling |
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 |