CN113065583A - Rectification process abnormity monitoring method based on online nonlinear discriminant feature analysis - Google Patents
Rectification process abnormity monitoring method based on online nonlinear discriminant feature analysis Download PDFInfo
- Publication number
- CN113065583A CN113065583A CN202110300826.XA CN202110300826A CN113065583A CN 113065583 A CN113065583 A CN 113065583A CN 202110300826 A CN202110300826 A CN 202110300826A CN 113065583 A CN113065583 A CN 113065583A
- Authority
- CN
- China
- Prior art keywords
- vector
- online
- formula
- matrix
- monitoring
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000012544 monitoring process Methods 0.000 title claims abstract description 38
- 238000004458 analytical method Methods 0.000 title claims abstract description 28
- 238000005070 sampling Methods 0.000 claims abstract description 34
- 230000002159 abnormal effect Effects 0.000 claims abstract description 29
- 239000013598 vector Substances 0.000 claims description 85
- 239000011159 matrix material Substances 0.000 claims description 22
- 238000012545 processing Methods 0.000 claims description 8
- 239000010779 crude oil Substances 0.000 claims description 6
- 239000000295 fuel oil Substances 0.000 claims description 6
- 239000003502 gasoline Substances 0.000 claims description 6
- 239000003350 kerosene Substances 0.000 claims description 6
- 238000010992 reflux Methods 0.000 claims description 6
- 239000000126 substance Substances 0.000 claims description 6
- 241000287196 Asthenes Species 0.000 claims description 3
- 239000000498 cooling water Substances 0.000 claims description 3
- 239000007788 liquid Substances 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000001960 triggered effect Effects 0.000 claims description 3
- 102000008297 Nuclear Matrix-Associated Proteins Human genes 0.000 claims description 2
- 108010035916 Nuclear Matrix-Associated Proteins Proteins 0.000 claims description 2
- 230000003321 amplification Effects 0.000 claims description 2
- 210000000299 nuclear matrix Anatomy 0.000 claims description 2
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 abstract description 5
- 238000000605 extraction Methods 0.000 abstract description 3
- 238000001311 chemical methods and process Methods 0.000 description 8
- 239000002283 diesel fuel Substances 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 238000012886 linear function Methods 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000035945 sensitivity Effects 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 238000009835 boiling Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000012880 independent component analysis Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000003209 petroleum derivative Substances 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Investigating Or Analyzing Materials Using Thermal Means (AREA)
- Monitoring And Testing Of Nuclear Reactors (AREA)
Abstract
The invention discloses a rectification process abnormity monitoring method based on online nonlinear discriminant feature analysis, which is used for carrying out nonlinear discriminant feature extraction on each online sampling data in real time through an online nonlinear discriminant feature analysis technology and aims to ensure that the characteristic extracted in real time can discriminate the hidden abnormal change characteristic in each sampling data to the maximum extent. The method has the following advantages: the method aims at maximizing the monitoring index, and can implement online nonlinear discriminant feature analysis on newly measured sampling data online in real time, so that features obtained after online analysis can deviate from zero points as much as possible; in this respect, the characteristic analysis technology implemented by the method is carried out on line, and the nonlinear abnormal change characteristic which can reflect the hidden nonlinear abnormal change in the online sampling data can be found out most, so that various abnormal working condition data in the rectification process can be effectively distinguished.
Description
Technical Field
The invention relates to a method for monitoring an abnormal state of a chemical process, in particular to a method for monitoring an abnormal state of a rectification process based on-line nonlinear discriminant feature analysis.
Background
The plate-type rectifying tower is widely used in the petrochemical industry to convert crude oil (or called as naphtha) into different petroleum products through different boiling points, which mainly comprises: gasoline, kerosene, light diesel, and fuel oil. In order to ensure the yield and purity of various products and reduce the production energy consumption, the rectification process is generally expected to be in a continuous and stable normal operation state. Therefore, the monitoring of the abnormity occurring in the operation of the rectification process in real time is the key for controlling the operation state of the rectification tower. Under the large background of current intelligent manufacturing, the utilization of chemical industry big data has become an intelligent metaphor. Therefore, the online monitoring of the abnormity of the rectification process according to the data-driven idea is the requirement of the development of the load era.
In the existing scientific and technical literature and patent documents, data-driven chemical process monitoring is developed according to the idea of single classification. On one hand, the chemical process is in a normal state of stable operation in most of time, and most of the acquired mass data belong to normal working conditions. On the other hand, the feature analysis algorithm suitable for unsupervised learning can mine the change features of the normal working condition data from a plurality of different angles. Therefore, on the premise of application of sampling data in an extremely small abnormal state, a feature analysis algorithm suitable for unsupervised learning is widely researched and applied in the technical field, and chemical process abnormality monitoring methods represented by principal component analysis and independent component analysis are endless.
However, due to the dynamic time sequence change characteristic of the chemical process, time sequence correlation exists between sampling data. Therefore, in addition to considering the cross correlation between different measurement variables of sample data, it is necessary to further relate to the timing correlation reflected in the sampling time sequence. However, from the perspective of monitoring abnormal conditions, the sampled data at different abnormal conditions may change in either cross correlation or timing correlation, or both. From the angle, the potential characteristics of the data of the normal working conditions are mined independently, so that the effectiveness and the sensitivity of the extracted characteristics for identifying the abnormal state cannot be guaranteed all the time. In addition, in consideration of the complex characteristics of the modern chemical process, the relation among different measurement variables can be found to be nonlinear through a differential equation, and the abnormal change of the nonlinear relation can also reflect the abnormal condition in the operation of the rectification process.
In summary, the rectification process is a physical separation device commonly used in chemical processes, and the stability and continuity of the operation state of the rectification process are self-evident in the petrochemical industry. In addition, the abnormal state of the operation of the rectification process can be reflected by the nonlinear relation characteristic between different measurement variables in the rectification process and the abnormal change of the time sequence characteristic in the sampling time sequence. Considering that different abnormal states can change different change characteristics in sample data, the effectiveness and the sensitivity of monitoring various abnormal states can be ensured only by implementing online characteristic extraction under the condition that real-time sampling data participates.
Disclosure of Invention
The invention aims to solve the main technical problems that: how to implement online nonlinear characteristic analysis on online sample data so that the characteristics extracted immediately can effectively judge the hidden abnormal change characteristics in the online sample data to achieve the purpose of efficiently monitoring the abnormal state in the rectification process. Specifically, the method of the invention carries out nonlinear discriminant feature extraction on each online sampling data in real time through an online nonlinear discriminant feature analysis technology, and aims to ensure that the characteristic extracted in real time can distinguish the latent abnormal change feature in each sampling data to the maximum extent.
The technical scheme adopted by the method for solving the problems is as follows: a rectification process abnormity monitoring method based on online nonlinear discriminant feature analysis comprises the following steps:
step (1): under the normal operation state of the plate-type rectifying tower, sequentially acquiring n sample data vectors x according to sampling time1，x2，…，xnEach sample data vector specifically includes m sample data, which is in turn: crude oil inlet flow, crude oil inlet temperature, top column pressure, top column outlet flow, top column reflux, gasoline outlet flow, kerosene outlet flow, light diesel outlet flow, fuel oil outlet flow, bottom column liquid outlet flow, bottom column reflux, condenser cooling water flow, reboiler steam flow, and temperatures of the trays in each layer; wherein the ith sample data vector xi∈Rm×1，Rm×1Representing a vector of real numbers in dimension m x 1, R representing a real numberSet, i ∈ {1, 2, …, n }.
If the number of tower plate layers in the plate-type rectifying tower is A, each tower plate layer is provided with a corresponding temperature measuring instrument, and the temperature data T of each tower plate layer is obtained in real time1，T2，…，TA. In addition, the product types in the middle of the tower specifically include gasoline, kerosene, light diesel oil and fuel oil. Thus, the outlet flow of each product in the middle of the column specifically comprises 4 flow data, namely: gasoline outlet flow F1Kerosene outlet flow F2Light diesel oil outlet flow F3And fuel oil outlet flow rate F4。
Step (2): using x1，x2，…，xnThe mean vector mu and the standard deviation vector delta of (d), respectively for x1，x2，…，xnCarrying out standardization processing to correspondingly obtain n data vectors
And (3): calculating a kernel matrix K epsilon R according to a formula shown in the specificationn×nRow i and column j element K (i, j):
wherein R isn×nA matrix of real numbers representing dimensions n x n, j ∈ {1, 2, …, n },
representation calculation
And
the distance between the two symbols, e represents a natural constant, alpha is a kernel parameter, and the upper label T represents a transposed symbol of the matrix or the vector.
And (4): k is subjected to centralization processing according to the formula shown below, so that a nuclear matrix is obtained
In the above formula, the matrix theta is formed by the element Rn×nAll elements in (1) are equal to 1.
And (5): N-D monitoring indices were calculated according to the following steps (5.1) to (5.4)
Wherein D represents the timing order, and the suggested value is D e {1, 2, …, 5 }.
Step (5.1): the initialization i is 1.
Step (5.2): according to the formula
The matrix Y of the constructed graph belongs to R(D+1)×n(ii) a Wherein the upper symbol T represents the transpose of a matrix or vector,
respectively representing kernel matrices
Column vector of the i + D, i + D-1, …, i column in (1).
Step (5.3): carrying out on-line nonlinear discriminant analysis on Y to obtain a left load vector u epsilon R(D+1)×1And the right load vector w ∈ Rn×1The specific implementation process is shown in the steps (A) to (D).
Step (A): the random initialization right load vector w is any real number vector of dimension n × 1.
Step (B): solving eigenvalue problem YwwTYTCalculating a left load vector u according to a formula u ═ p/| | p | | | in the characteristic vector p corresponding to the maximum characteristic value η in the equation p ═ η p; wherein the content of the first and second substances,
indicating the length of the computed feature vector p.
Step (C): solving eigenvalue problems
In the formula, the eigenvector q corresponding to the maximum eigenvalue lambda is calculated according to the formula
And calculating to obtain a right load vector w.
Step (D): judging whether w is converged; if not, returning to the step (B); if yes, obtaining a left load vector u epsilon R(D +1)×1And the right load vector w ∈ Rn×1。
Step (5.4): according to the formula
Calculating the ith monitoring index
Then, judging whether the condition i is less than N; if yes, setting i to i +1 and returning to the step (5.2); if not, obtaining
And (6): will be provided with
The maximum value in (1) is recorded as
Then according to the formula
Calculating to obtain the upper control limit
Wherein γ represents an amplification factor, and its value isThe range is gamma e [1.2, 1.4]。
It should be noted that the steps (a) to (D) are specific implementation processes of the online discriminant feature analysis technique according to the present invention. In fact, the online discriminant feature analysis technique aims to satisfy the objective function as shown below:
wherein φ (X) represents a matrix
Mapping to phi (X) in a high dimensional space by a non-linear function phi, phi (Z) representing the matrix
Mapped to phi (Z) in the high dimensional space by a non-linear function phi. The kernel learning concept is characterized in that the specific form of the non-linear function phi need not be known, but only the corresponding kernel matrix K ═ phi (X)T。
In addition, since it is necessary to center φ (X) so that the mean value is 0, it is common to center K to obtain a kernel matrix
And then carrying out subsequent calculation operation. According to the nuclear learning idea, the vector v can be obtained by calculating the right load vector w, namely: v ═ phi (X)Tw. Therefore, equation c above can be equivalently transformed into an objective function as shown below:
the solution of the above equation (r) may use the classical langrangian multiplier method, i.e. construct the langrangian function L as shown below by the lagrangian multipliers λ and η.
Consider | | uTYw||2＝(uTYw)2＝uTYwwTYTu＝wTYTuuTYw, then the partial differential equation shown below is obtained:
this in turn translates into two eigenvalue problems as shown below:
YtvvTYt Tu＝ηu ⑦
in addition, if u is multiplied on both sides of the equations of the above formula (c) and (b), respectivelyTAnd wTThen λ ═ η ═ u can be obtainedTYw||2. Therefore, the maximum eigenvalues in equations (c) and (b) need to be solved, and the corresponding eigenvectors need to be length constrained, i.e.: u. ofTu is 1 and
and (7): sample data vector x of latest sampling moment is acquired onlinet∈Rm×1And performing the same normalization process as in step (2) to obtain a data vector
Where the subscript t denotes the latest sampling instant.
And (8): the kernel vector k is calculated according to the formula shown belowt∈Rn×1The ith element k int(i)：
And (9): according to the formula shown below for ktPerforming centralization processing to obtain kernel vector
Wherein the vector theta ∈ Rn×1Each element in (1) is equal to 1.
Step (10): building a graph matrix
And set Y ═ YtThen, carrying out online nonlinear discriminant feature analysis according to the steps (A) to (D) to obtain a left load vector u and a right load vector w; wherein the content of the first and second substances,
respectively represents kt，kt-1…，kt-DThe kernel vector after the centering, and kt，kt-1…，kt-DRepresenting the kernel vectors at t, t-1, …, t-D sample times, respectively.
Step (11): according to the formula
Calculating a monitoring index
Then, whether the condition is satisfied is judged
If so, the rectifying tower operates normally at the current sampling moment, and the step (7) is returned to continue to utilize the sample data vector at the latest sampling moment to implement abnormal monitoring; if not, executing step (12).
Step (12): returning to step (7) to continue using the latest sampling timeCarrying out abnormity monitoring on the sample data vectors, and if the monitoring indexes of 6 continuous latest sampling moments all meet the condition
An abnormal state alarm is triggered.
By carrying out the steps described above, the advantages of the method of the invention are presented below.
The method has the following advantages: the method is different from the traditional chemical process abnormity monitoring method based on the nuclear learning method, and implements online nonlinear discriminant feature analysis on newly measured sampling data online in real time, aiming at maximizing the monitoring index, so that the features obtained after online analysis can deviate from the zero point as much as possible; in this respect, the characteristic analysis technology implemented by the method is carried out on line, and the nonlinear abnormal change characteristic which can reflect the hidden nonlinear abnormal change in the on-line sampling data can be found out most, so that various abnormal working condition data can be effectively distinguished.
Drawings
FIG. 1 is a schematic flow chart of the method of the present invention.
Fig. 2 is a schematic diagram of a production process of a plate-type rectifying column.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
The invention discloses a rectification process abnormity monitoring method based on online nonlinear discriminant feature analysis, wherein the implementation process of the most key online nonlinear discriminant feature analysis technology of the method is shown in figure 1. The following describes a specific embodiment of the method of the present invention in conjunction with a specific application example.
The schematic diagram of the production process of the rectifying tower in this embodiment is shown in fig. 2. In this embodiment, 1023 sample data of the plate-type rectifying tower under the normal operation condition are collected according to the time sequence, that is, n is 1023, and the following steps (1) to (6) are performed, so that each parameter required for performing online anomaly monitoring is obtained.
Step (1): under the normal operation state of the plate-type rectifying tower, sequentially acquiring n sample data vectors x according to sampling time1，x2，…，xnEach sample data vector specifically includes m sample data, which is in turn: crude oil inlet flow FinAnd inlet temperature TinPressure at the top of the column PTopOutlet flow rate FTopAnd a reflux amount BTopGasoline outlet flow F1Kerosene outlet flow F2Outlet flow F of light diesel oil3Outlet flow rate of fuel oil F4Liquid outlet flow F at the bottom of the columnLWith the amount of reflux BLTemperature T of A-stage trays1，T2，…，TACondenser cooling water flow rate FCReboiler steam flow FS。
Step (2): using x1，x2，…，xnThe mean vector mu and the standard deviation vector delta of (d), respectively for x1，x2，…，xnCarrying out standardization processing to correspondingly obtain n data vectors
And (3): calculating a kernel matrix K epsilon R according to the formulan×nRow i and column j elements K (i, j).
And (4): according to the formula II, K is subjected to centralization treatment, and thus a kernel matrix is obtained
And (5): after the time sequence order D is set to 3, N-D monitoring indexes are calculated according to the steps (5.1) to (5.4)
And (6): will be provided with
The maximum value in (1) is recorded as
Then according to the formula
Calculating to obtain the upper control limit
Wherein γ is 1.3.
After the steps (1) to (6) are completed, each parameter required for implementing online anomaly monitoring can be obtained, and the method specifically comprises the following steps: the mean vector mu and the standard deviation vector delta in the step (2), and the kernel matrix K belonging to the R in the step (3)n×nThe order of the timing sequence in step (5), and the upper limit of control in step (6)
Then, the online anomaly monitoring for the plate-type rectifying tower can be continuously implemented according to the sample data vector of the latest sampling moment according to the following steps (7) to (12).
And (7): sample data vector x of latest sampling moment is acquired onlinet∈Rm×1And performing the same normalization process as in step (2) to obtain a data vector
And (8): ninthly, calculating a nuclear vector k according to the formulat∈Rn×1The ith element k int(i)。
And (9): according to the above formula r to ktPerforming centralization processing to obtain kernel vector
Step (10): building a graph matrix
And set Y ═ YtThen, on-line nonlinear discriminant feature analysis is performed according to the following steps (a) to (D), thereby obtaining a left load vector u and a right load vector w.
Step (A): the random initialization right load vector w is any real number vector of dimension n × 1.
Step (B): solving eigenvalue problem YwwTYTCalculating a left load vector u according to a formula u ═ p/| | p | | | in the characteristic vector p corresponding to the maximum characteristic value η in the equation p ═ η p; wherein the content of the first and second substances,
step (C): solving eigenvalue problems
In the formula, the eigenvector q corresponding to the maximum eigenvalue lambda is calculated according to the formula
And calculating to obtain a right load vector w.
Step (D): judging whether w is converged; if not, returning to the step (B); if yes, obtaining a left load vector u epsilon R(D +1)×1And the right load vector w ∈ Rn×1。
Step (11): according to the formula
Calculating a monitoring index
Then, whether the condition is satisfied is judged
If so, the rectifying tower operates normally at the current sampling moment, and the step (7) is returned to continue to utilize the sample data vector at the latest sampling moment to implement abnormal monitoring; if not, executing step (12).
Step (12): returning to the step (7) to continue utilizing the sample data at the latest sampling momentPerforming abnormal monitoring if the conditions are satisfied for 6 times
An abnormal state alarm is triggered.
Claims (2)
1. A rectification process abnormity monitoring method based on online nonlinear discriminant feature analysis is characterized by comprising the following steps:
step (1): under the normal operation state of the plate-type rectifying tower, sequentially acquiring n sample data vectors x according to sampling time1，x2，…，xnEach sample data vector specifically includes m sample data, which is in turn: crude oil inlet flow, crude oil inlet temperature, top column pressure, top column outlet flow, top column reflux, gasoline outlet flow, kerosene outlet flow, light diesel outlet flow, fuel oil outlet flow, bottom column liquid outlet flow, bottom column reflux, condenser cooling water flow, reboiler steam flow, and temperatures of the trays in each layer; wherein the ith sample data vector xi∈Rm×1，Rm×1Representing a real number vector of dimension m × 1, R representing a real number set, i ∈ {1, 2, …, n };
step (2): using x1，x2，…，xnThe mean vector mu and the standard deviation vector delta of (d), respectively for x1，x2，…，xnCarrying out standardization processing to correspondingly obtain n data vectors
And (3): calculating a kernel matrix K epsilon R according to a formula shown in the specificationn×nRow i and column j element K (i, j):
wherein R isn×nA matrix of real numbers representing dimensions n x n,
e represents a natural constant, alpha is a kernel parameter, and the upper label T represents a transposed symbol of a matrix or a vector;
and (4): k is subjected to centralization processing according to the formula shown below, so that a nuclear matrix is obtained
In the above formula, the matrix theta is formed by the element Rn×nAll elements in (1);
and (5): N-D monitoring indices were calculated according to the following steps (5.1) to (5.4)
Wherein D represents a time sequence order;
step (5.1): initializing i to 1;
step (5.2): according to the formula
The matrix Y of the constructed graph belongs to R(D+1)×n(ii) a Wherein the upper symbol T represents the transpose of a matrix or vector,
respectively representing kernel matrices
Column vector of the i + D, i + D-1, …, i column in (1);
step (5.3): carrying out on-line nonlinear discriminant analysis on Y to obtain a left load vector u epsilon R(D+1)×1And the right load vector w ∈ Rn×1；
Step (5.4): according to the formula
Calculating the ith monitoring index
Then, judging whether the condition i is less than N; if yes, setting i to i +1 and returning to the step (5.2); if not, obtaining
And (6): will be provided with
The maximum value in (1) is recorded as
Then according to the formula
Calculating to obtain the upper control limit
Wherein γ represents an amplification factor;
and (7): sample data vector x of latest sampling moment is acquired onlinet∈Rm×1And performing the same normalization process as in step (2) to obtain a data vector
Wherein the subscript t denotes the latest sampling time;
and (8): the kernel vector k is calculated according to the formula shown belowt∈Rn×1The ith element k int(i)：
And (9): according to the formula shown below for ktPerforming centralization processing to obtain kernel vector
Wherein the vector theta ∈ Rn×1Each element in (1);
step (10): building a graph matrix
And set Y ═ YtThen, carrying out online nonlinear discriminant feature analysis to obtain a left load vector u and a right load vector w; wherein the content of the first and second substances,
respectively represents kt，kt-1…，kt-DThe kernel vector after the centering, and kt，kt-1…，kt-DRespectively representing kernel vectors at t, t-1, … and t-D sampling moments;
step (11): according to the formula
Calculating a monitoring index
Then, whether the condition is satisfied is judged
If so, the rectifying tower operates normally at the current sampling moment, and the step (7) is returned to continue to utilize the sample data vector at the latest sampling moment to implement abnormal monitoring; if not, executing the step (12);
step (12): return toReturning to the step (7) to continue to use the sample data vector at the latest sampling moment to implement anomaly monitoring, if the monitoring indexes of 6 continuous latest sampling moments all meet the condition
An abnormal state alarm is triggered.
2. The rectification process anomaly monitoring method based on online nonlinear discriminant feature analysis according to claim 1, wherein the specific implementation process of implementing online nonlinear discriminant feature analysis in the step (5.3) and the step (10) is as follows:
step (A): randomly initializing a right load vector w to be any real number vector of n multiplied by 1 dimension;
step (B): solving eigenvalue problem YwwTYTCalculating a left load vector u according to a formula u ═ p/| | p | | | in the characteristic vector p corresponding to the maximum characteristic value η in the equation p ═ η p; wherein the content of the first and second substances,
step (C): solving eigenvalue problems
In the formula, the eigenvector q corresponding to the maximum eigenvalue lambda is calculated according to the formula
Calculating to obtain a right load vector w;
step (D): judging whether w is converged; if not, returning to the step (B); if yes, obtaining a left load vector u epsilon R(D+1)×1And the right load vector w ∈ Rn×1。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110300826.XA CN113065583B (en) | 2021-03-13 | 2021-03-13 | Rectification process abnormity monitoring method based on online nonlinear discriminant feature analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110300826.XA CN113065583B (en) | 2021-03-13 | 2021-03-13 | Rectification process abnormity monitoring method based on online nonlinear discriminant feature analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113065583A true CN113065583A (en) | 2021-07-02 |
| CN113065583B CN113065583B (en) | 2023-11-14 |
Family
ID=76562865
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110300826.XA Active CN113065583B (en) | 2021-03-13 | 2021-03-13 | Rectification process abnormity monitoring method based on online nonlinear discriminant feature analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113065583B (en) |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2008056638A1 (en) * | 2006-11-06 | 2008-05-15 | Fujifilm Ri Pharma Co., Ltd. | Brain image diagnosis supporting method, program, and recording method |
| CN101714273A (en) * | 2009-05-26 | 2010-05-26 | 北京银丰新融科技开发有限公司 | Rule engine-based method and system for monitoring exceptional service of bank |
| CN103048041A (en) * | 2012-12-20 | 2013-04-17 | 北京信息科技大学 | Fault diagnosis method of electromechanical system based on local tangent space and support vector machine |
| CN103400021A (en) * | 2013-06-06 | 2013-11-20 | 浙江工业大学 | KDSNE hydroelectric generating set noise source detection method based on heavy tailed distribution |
| CN105334504A (en) * | 2015-09-08 | 2016-02-17 | 西安电子科技大学 | Radar target identification method based on large-boundary nonlinear discrimination projection model |
| CN109538311A (en) * | 2018-09-21 | 2019-03-29 | 浙江大学 | Control performance method of real-time towards steam turbine in high-end power generating equipment |
| CN109919130A (en) * | 2019-03-21 | 2019-06-21 | 西安交通大学 | A kind of quality control chart mode identification method based on improved adaptive GA-IAGA optimization |
| CN111914889A (en) * | 2020-06-13 | 2020-11-10 | 宁波大学 | Rectifying tower abnormal state identification method based on brief kernel principal component analysis |
| CN111914886A (en) * | 2020-06-13 | 2020-11-10 | 宁波大学 | Nonlinear chemical process monitoring method based on online brief kernel learning |
| CN111914214A (en) * | 2020-06-13 | 2020-11-10 | 宁波大学 | PTA production process soft measurement method based on reduced KPLS model |
| CN111905396A (en) * | 2020-06-13 | 2020-11-10 | 宁波大学 | Rectification process real-time monitoring method based on online sampling data driving |
| CN111913447A (en) * | 2020-06-13 | 2020-11-10 | 宁波大学 | Online data-driven nonlinear chemical process monitoring method |
| CN111913460A (en) * | 2019-05-20 | 2020-11-10 | 宁波大学 | Fault monitoring method based on sequence correlation local preserving projection algorithm |
-
2021
- 2021-03-13 CN CN202110300826.XA patent/CN113065583B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2008056638A1 (en) * | 2006-11-06 | 2008-05-15 | Fujifilm Ri Pharma Co., Ltd. | Brain image diagnosis supporting method, program, and recording method |
| CN101714273A (en) * | 2009-05-26 | 2010-05-26 | 北京银丰新融科技开发有限公司 | Rule engine-based method and system for monitoring exceptional service of bank |
| CN103048041A (en) * | 2012-12-20 | 2013-04-17 | 北京信息科技大学 | Fault diagnosis method of electromechanical system based on local tangent space and support vector machine |
| CN103400021A (en) * | 2013-06-06 | 2013-11-20 | 浙江工业大学 | KDSNE hydroelectric generating set noise source detection method based on heavy tailed distribution |
| CN105334504A (en) * | 2015-09-08 | 2016-02-17 | 西安电子科技大学 | Radar target identification method based on large-boundary nonlinear discrimination projection model |
| CN109538311A (en) * | 2018-09-21 | 2019-03-29 | 浙江大学 | Control performance method of real-time towards steam turbine in high-end power generating equipment |
| CN109919130A (en) * | 2019-03-21 | 2019-06-21 | 西安交通大学 | A kind of quality control chart mode identification method based on improved adaptive GA-IAGA optimization |
| CN111913460A (en) * | 2019-05-20 | 2020-11-10 | 宁波大学 | Fault monitoring method based on sequence correlation local preserving projection algorithm |
| CN111914889A (en) * | 2020-06-13 | 2020-11-10 | 宁波大学 | Rectifying tower abnormal state identification method based on brief kernel principal component analysis |
| CN111914886A (en) * | 2020-06-13 | 2020-11-10 | 宁波大学 | Nonlinear chemical process monitoring method based on online brief kernel learning |
| CN111914214A (en) * | 2020-06-13 | 2020-11-10 | 宁波大学 | PTA production process soft measurement method based on reduced KPLS model |
| CN111905396A (en) * | 2020-06-13 | 2020-11-10 | 宁波大学 | Rectification process real-time monitoring method based on online sampling data driving |
| CN111913447A (en) * | 2020-06-13 | 2020-11-10 | 宁波大学 | Online data-driven nonlinear chemical process monitoring method |
Non-Patent Citations (2)
| Title |
|---|
| 李巍华, 史铁林, 杨叔子: "基于非线性判别分析的故障分类方法研究", 振动工程学报, no. 02, pages 231 - 237 * |
| 肖应旺;张绪红;陈呈国;: "一种新的非线性FDA间歇过程监控方法", 计算机与应用化学, no. 07, pages 29 - 35 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113065583B (en) | 2023-11-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yuan et al. | A dynamic CNN for nonlinear dynamic feature learning in soft sensor modeling of industrial process data | |
| Jang et al. | Adversarial autoencoder based feature learning for fault detection in industrial processes | |
| CN106845796B (en) | One kind is hydrocracked flow product quality on-line prediction method | |
| CN110009020B (en) | Nonlinear process monitoring method based on multi-core principal component analysis model | |
| CN108469805B (en) | Distributed dynamic process monitoring method based on dynamic optimal selection | |
| CN108549908A (en) | Chemical process fault detection method based on more sampled probability core principle component models | |
| CN111914889A (en) | Rectifying tower abnormal state identification method based on brief kernel principal component analysis | |
| CN112613536A (en) | Near infrared spectrum diesel grade identification method based on SMOTE and deep learning | |
| CN111914887B (en) | Novel multi-mode chemical process abnormal state detection method | |
| Liu et al. | Deep nonlinear dynamic feature extraction for quality prediction based on spatiotemporal neighborhood preserving SAE | |
| CN110362063B (en) | Fault detection method and system based on global maintenance unsupervised kernel extreme learning machine | |
| CN110490496B (en) | Method for screening sensitive variables influencing product quality in complex industrial process based on stepwise reduction | |
| CN110942258B (en) | Performance-driven industrial process anomaly monitoring method | |
| CN116821695B (en) | Semi-supervised neural network soft measurement modeling method | |
| CN113065583A (en) | Rectification process abnormity monitoring method based on online nonlinear discriminant feature analysis | |
| CN111913415B (en) | Continuous stirring reaction kettle operation state monitoring method based on time sequence data analysis | |
| CN111912638B (en) | Rectifying tower fault diagnosis method for online fault source identification | |
| Yuan et al. | Variable correlation analysis-based convolutional neural network for far topological feature extraction and industrial predictive modeling | |
| CN112149054B (en) | Construction and application of orthogonal neighborhood preserving embedding model based on time sequence expansion | |
| Xu et al. | Similarity and sparsity collaborative embedding and its application to robust process monitoring | |
| CN111914888A (en) | Chemical process monitoring method integrating multi-working-condition identification and fault detection | |
| CN114439457A (en) | Method and system for evaluating health state of rod-pumped well | |
| CN111914886A (en) | Nonlinear chemical process monitoring method based on online brief kernel learning | |
| CN111914214B (en) | PTA production process soft measurement method based on reduced KPLS model | |
| Lyu et al. | Image-based process monitoring using deep belief networks |
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 |