CN111914888A - Chemical process monitoring method integrating multi-working-condition identification and fault detection - Google Patents
Chemical process monitoring method integrating multi-working-condition identification and fault detection Download PDFInfo
- Publication number
- CN111914888A CN111914888A CN202010570615.3A CN202010570615A CN111914888A CN 111914888 A CN111914888 A CN 111914888A CN 202010570615 A CN202010570615 A CN 202010570615A CN 111914888 A CN111914888 A CN 111914888A
- Authority
- CN
- China
- Prior art keywords
- matrix
- condition
- working
- vector
- chemical process
- 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.)
- Pending
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 57
- 238000001311 chemical methods and process Methods 0.000 title claims abstract description 53
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000001514 detection method Methods 0.000 title claims abstract description 20
- 238000004519 manufacturing process Methods 0.000 claims abstract description 48
- 238000005070 sampling Methods 0.000 claims abstract description 29
- 230000008569 process Effects 0.000 claims abstract description 13
- 239000011159 matrix material Substances 0.000 claims description 89
- 239000013598 vector Substances 0.000 claims description 77
- 230000009466 transformation Effects 0.000 claims description 13
- 230000002159 abnormal effect Effects 0.000 claims description 10
- 239000000126 substance Substances 0.000 claims description 6
- 230000001960 triggered effect Effects 0.000 claims description 5
- 238000009826 distribution Methods 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 2
- 125000004432 carbon atom Chemical group C* 0.000 claims description 2
- 239000000470 constituent Substances 0.000 claims description 2
- 238000006243 chemical reaction Methods 0.000 abstract description 4
- 238000003756 stirring Methods 0.000 abstract description 2
- 238000005516 engineering process Methods 0.000 description 5
- 238000000513 principal component analysis Methods 0.000 description 4
- 238000011160 research Methods 0.000 description 3
- 239000000498 cooling water Substances 0.000 description 2
- 238000005265 energy consumption Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 102000002274 Matrix Metalloproteinases Human genes 0.000 description 1
- 108010000684 Matrix Metalloproteinases Proteins 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- FFBHFFJDDLITSX-UHFFFAOYSA-N benzyl N-[2-hydroxy-4-(3-oxomorpholin-4-yl)phenyl]carbamate Chemical compound OC1=C(NC(=O)OCC2=CC=CC=C2)C=CC(=C1)N1CCOCC1=O FFBHFFJDDLITSX-UHFFFAOYSA-N 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 238000012824 chemical production Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000000491 multivariate analysis Methods 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 239000000376 reactant Substances 0.000 description 1
- 238000010992 reflux Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000012549 training 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
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/211—Selection of the most significant subset of features
- G06F18/2113—Selection of the most significant subset of features by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
-
- 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
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
- G05B23/0224—Process history based detection method, e.g. whereby history implies the availability of large amounts of data
-
- 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
Abstract
The invention discloses a chemical process monitoring method integrating multi-working-condition identification and fault detection, which solves the main problems of two aspects: firstly, how to identify multiple working conditions on the premise that the number of normal production working conditions is not definite; secondly, after the attribution of each working condition data is identified, how to further mine the difference among the multi-working condition sampling data, and a fault detection model is established. The method is a multi-working-condition chemical process state monitoring method integrating multi-working-condition identification and fault detection, and the implementation process of the related multi-working-condition identification does not need to know the total number of normal production working conditions in advance. In addition, the online state monitoring process related by the method does not need to judge which production working condition the current sampling data belongs to online in real time. Finally, the effectiveness of the method in monitoring the operating state of the multi-working-condition continuous stirring reaction kettle can be fully illustrated through specific implementation cases.
Description
Technical Field
The invention relates to a method for monitoring the running state of a chemical process, in particular to a method for monitoring the chemical process integrating multi-working-condition identification and fault detection.
Background
Due to the rapid development of the instrumentation technology and the wide application of the field bus technology, the modern chemical process usually acquires data information such as temperature, pressure, flow rate and the like in the production process through the instrumentation and measurement instrument uninterruptedly. The mass-stored sampling data provides a solid data base for the construction of an intelligent chemical system, for example, the data is used for monitoring the operation state of a chemical process so as to ensure the production safety and the product quality stability. In recent decades, data-driven chemical process state monitoring receives more and more attention in the field of safety chemical production, and a large amount of manpower and financial research and novel data-driven chemical process state monitoring technology is applied in both academic circles and industrial circles. As the number of the chemical process measurement variables is more, a plurality of multivariate statistical process monitoring methods based on multivariate analysis algorithms appear in the data-driven chemical process monitoring. The core of these mainstream process monitoring method implementations is primarily concerned with the mining of potential features of the data. In other words, the data-driven models that are built are all intended to extract features that are hidden in the sampled data.
A typical core production unit of a chemical process object comprises: reaction equipment, separation equipment, cooling equipment and a reflux device. The devices are linked with each other, and the operation mechanism and the control system loop are complex. In addition, due to the reasons of fluctuating order requirements, energy consumption scheduling problems and the like of chemical plants, chemical process objects do not always operate under a certain fixed working condition. Even if various abnormal conditions (such as actuator faults, sensor faults and the like) are not considered, the working condition that the chemical process maintains the normal operation state can be in multiple production modes. For example, due to high energy consumption of chemical enterprises, the output is usually adjusted according to fluctuation of electricity price or market price of other raw materials, so that the chemical process has a plurality of normal production conditions. Implementing data-driven condition monitoring for multi-regime chemical processes generally involves two tasks: firstly, how to identify a plurality of production working conditions of a chemical process from sampling data in an off-line modeling stage; secondly, how to implement online fault detection for the chemical process under multiple working conditions. Therefore, a method technology integrating multi-working condition identification and fault detection is needed for state monitoring of a multi-working condition chemical process.
In the existing scientific research documents and patent technical materials, research aiming at monitoring the state of a multi-working-condition chemical process focuses on how to establish a corresponding fault detection model for sampling data under each production working condition. For example, the most classical approaches are: and respectively establishing corresponding fault detection models for the sampling data of each working condition by using a Principal Component Analysis (PCA) algorithm, and finally respectively performing online fault detection on the multi-working-condition chemical process by using the multi-working-condition PCA model. A significant problem of the classical multi-working-condition PCA method technology is that: the difference between the sampled data under each production condition is not considered. In the problem of multi-condition identification, it is usually assumed that the condition to which the sampled data belongs is known or is determined by a clustering algorithm. For example, the C-means clustering algorithm is the most and most classical method applied. However, the C-means clustering algorithm has an obvious technical shorthand, that is, the number of working conditions in the multi-working-condition training data needs to be predicted, and the C-means is easy to fall into local minimum, so that the multi-working-condition identification fails. Therefore, the invention provides a more feasible multi-working-condition chemical process state monitoring method integrating multi-working-condition identification and fault detection into a whole, which is imperative!
Disclosure of Invention
The invention aims to solve the main technical problems that: the invention provides a state monitoring method integrating multi-working-condition state identification and fault detection for a multi-working-condition chemical process. Specifically, the main problem solved involves two aspects: firstly, how to identify multiple working conditions on the premise that the number of normal production working conditions is not definite; secondly, after the attribution of each working condition data is identified, how to further mine the difference among the multi-working condition sampling data, and a fault detection model is established.
The technical scheme adopted by the method for solving the problems is as follows: a chemical process monitoring method integrating multi-working-condition identification and fault detection comprises the following steps.
Step (1): collecting N sample data x under normal operation state of chemical process by using measuring instrument in chemical process object1，x2，…，xNWherein x isi∈Rm×1Represents the ith sample data, Rm×1Real number vector representing m x 1 dimension, m being total number of measured variables of the chemical process objectThe subscript i ═ 1, 2, …, N.
Considering that a plurality of production working conditions can occur in the chemical process, the switching between the production working conditions is not too frequent. In other words, under a certain production condition, the chemical process can be continuously and stably operated for a long time. Thus, the N sample data x1，x2，…，xNFirst n sample data x in (1)1，x2，…，xnThe theory belongs to the same production working condition.
Step (2): determining the first n sample data x1，x2，…，xnAfter belonging to the same production working condition, according to the formula mu0＝(x1+x2+…+xn) N and the following formula (i) calculate the mean vector mu separately0And the standard deviation vector sigma0：
Wherein N is less than N and (x)j-μ0)·(xj-μ()) Represents a vector (x)j-μ0) And vector (x)j-μ0) Where the elements in the same position are multiplied by a subscript number j ═ 1, 2, …, n.
And (3): using mean vector mu0Sum standard deviation vector σ0For matrix X ═ X1 x2 … xN]∈Rm×NImplementing a normalization process to obtain a new matrix
And will be
The column vectors of the first n columns form a reference matrix X0∈Rm×nWill be
The column vectors of the middle and rear N-N columns form a test matrix Y0∈Rm×(N-n)WhereinRm×nRepresenting a matrix of real numbers in m x n dimensions, Rm×(N-n)A matrix of real numbers representing dimensions m × (N-N).
And (4): identifying the N sample data x acquired in step (1) by using the following steps (4.1) to (4.7)1，x2，…，xNAnd sequentially marking the plurality of identified production conditions as 1, 2, …, C, wherein C is the total number of the identified production conditions.
Step (4.1): according to the formula
Calculating a symmetric matrix phi epsilon Rm×mWherein A is-1/2And after representing the inverse matrix of the calculation matrix A, carrying out root number opening processing, namely: (A)-1/2)(A-1/2)＝A-1。
Step (4.2): calculating eigenvectors corresponding to all eigenvalues of the symmetric matrix phi, and arranging all eigenvalues in descending order according to magnitude to obtain lambda1≥λ2≥…≥λmCharacteristic value lambda1，λ2，…，λmThe corresponding characteristic vectors are v in sequence1，v2，…，vm。
Step (4.3): according to the characteristic value lambda1，λ2，…，λmDetermining the parameter D0。
Step (4.4): will D0A feature vector
Composition feature matrix
Then, the transformation matrix is calculated
Step (4.5): according to the formula
Calculating score matrix S0Then, calculate S again0Covariance matrix of
Step (4.6): according to the formula
Calculating a monitoring index vector psi0∈RN×1Then, drawing psi0In which
diag { } denotes an operation of converting a matrix diagonal element in braces into a column vector.
Step (4.7): according to the variation curve drawn in the step (4.6), a plurality of production working conditions are identified and marked as 1, 2, … and C in sequence, and N sample data x collected in the step (1) are determined1，x2，…，xNBelonging to production working conditions.
And (5): calculating a mean vector mu and a standard deviation vector sigma of all column vectors in the matrix X, and normalizing the data matrix X by using the mu and the sigma to obtain a matrix
And (6): according to the production working condition attribution of each sample data determined in the step (4), the matrix is processed
Divided into C data matrices
Wherein the content of the first and second substances,
is a matrix under the k-th production condition,
represents mxNkReal number matrix of dimension, NkThe total number of sample data attributed to the k-th production run, k is 1, 2, …, C.
It is worth pointing out that, since these C matrices are used
Is a slave matrix
Is separated out so that N is equal to N1+N2+…+NC。
And (7): determining a main transformation matrix W epsilon R according to the steps (7.1) to (7.4) shown as followsm×DAnd a sub-transform matrix
Step (7.1): calculating the symmetry matrix theta epsilon R according to the formula shown belowm×m：
In the above formula II, matrix
Is composed of a matrix
Removing matrix
The remaining C-1 matrices are combined.
Step (7.2): calculating eigenvectors corresponding to m eigenvalues in the symmetric matrix theta, and arranging the m eigenvalues in descending order according to the magnitude to obtain eta1≥η2≥…≥ηmCharacteristic value eta1，η2，…，ηmThe corresponding characteristic vectors are p in sequence1，p2，…，pm。
Step (7.3): according to the characteristic value eta1，η2，…，ηmDetermines the parameter D.
Step (7.4): the first D feature vectors p1，p2，…，pDThe constituent master feature matrix P ═ P1，p2，…，pD]The last m-D eigenvectors pD+1，pD+2，…，pmComposing a sub-feature matrix
Step (7.5): calculating main transformation matrix
And a sub-transform matrix
And (8): according to the formula
Calculating a dominance score matrix SkThen, calculate S againkMean vector mu of all column vectors inkAnd SkOf the covariance matrix Λk。
And (9): the upper control limit U is determined according to the formula shown belowlimAnd control upper limit Qlim：
In the above formula, the first and second carbon atoms are,
the chi-square distribution with D degree of freedom is distributed at the confidence level alphaThe value of the compound under the condition is,
representing a degree of freedom of
The value of the chi-square distribution under the condition of the confidence degree alpha,
step (10): keeping mean vector sum mu, standard deviation vector sigma, main transformation matrix W and auxiliary transformation matrix
Mean vector mu1，μ2，…，μCCovariance matrix Λ1，Λ2，…，ΛCAnd an upper control limit UlimAnd QlimAnd the real-time calling is carried out when the online process monitoring is implemented.
The steps (1) to (10) are the off-line modeling stage of the method, and after the off-line modeling stage is finished, the sample data x measured in real time can be utilizedt∈Rm×1And monitoring the operating state of the multi-working-condition chemical process.
Step (11): sample data x of latest sampling moment of multi-condition chemical process object is acquired on linet∈Rm×1And using the mean vector sum mu and the standard deviation vector sigma to xtCarrying out standardization to obtain vector
Where the lower reference sign t denotes the latest sampling instant.
Step (12): according to the formula respectively
And
calculating a principal score vector stAnd the sub-score vector
Then, according to formula Ak＝(st-μk)TΛk(st-μk) Calculating the Mahalanobis squared distance AkWhere k is 1, 2, …, C.
Step (13): respectively calculating the monitoring indexes U according to the formula shown in the specificationtAnd a monitoring index Qt：
Ut＝min{A1，A2，…，AC} ⑤
In the above formula, min { A }1，A2，…，ACDenotes A1，A2，…，ACThe medium minimum value.
Step (14): judging whether the condition U is satisfiedt≤UlimAnd Qt≤Qlim(ii) a If yes, the chemical process does not have an abnormal state during operation at the current sampling moment, and the step (11) is returned to continue to monitor the state of the next latest sampling moment; if not, step (15) is executed to decide whether an abnormal state occurs.
Step (15): returning to the step (11) to continue to carry out state monitoring on the next latest sampling moment, and if the monitoring indexes of 3 continuous sampling moments do not meet the judgment condition in the step (14), the production process breaks down and a fault alarm is triggered; otherwise, an abnormal state alarm is not triggered, and the step (11) is returned to continue to carry out state monitoring on the next latest sampling moment.
Compared with the traditional multi-working-condition chemical process state monitoring method, the method has the following advantages.
The method is a multi-working-condition chemical process state monitoring method integrating multi-working-condition identification and fault detection, and the implementation process of the related multi-working-condition identification does not need to know the total number of normal production working conditions in advance. In addition, the online state monitoring process related by the method does not need to judge which production working condition the current sampling data belongs to online in real time. Finally, the effectiveness of the method of the present invention in monitoring the operating state of a multi-condition continuous stirred tank reactor can be fully demonstrated by the following specific embodiments.
Drawings
FIG. 1 is a schematic flow chart of the method of the present invention.
FIG. 2 is a schematic diagram of a continuous stirred tank reactor.
FIG. 3 shows a monitoring indicator vector psi0Graph of the variation of (c).
FIG. 4 is a real-time graph of the method of the present invention monitoring actuator failure.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
As shown in fig. 1, the present invention discloses a chemical process monitoring method integrating multi-condition identification and fault detection, and a specific embodiment of the method of the present invention is described below with reference to a specific application example.
As shown in FIG. 2, a flow chart of a Continuous Stirred Tank Reactor (CSTR) and its corresponding measuring instrument are shown. The CSTR production unit is the most common production facility in a chemical plant, and the application in this embodiment is a CSTR facility that involves an exothermic reaction process. Therefore, the CSTR equipment is equipped with a condenser to reduce the temperature of the reactant outlet. As can be seen from fig. 2, m is 7 measured variables related to the continuous stirring reaction kettle, which are respectively: feed flow, reactor pressure, reactor liquid level, reactor temperature, reactor feed valve opening, reactor condensate flow, and condenser cooling water flow.
Due to the yield scheduling reason, the feed flow of the CSTR equipment can be correspondingly adjusted according to different requirements of the yield, so that the chemical process object can normally run under a plurality of production working conditions and is a typical multi-working-condition chemical process object.
Firstly, the off-line modeling stage of the method is implemented by using sample data acquired by a CSTR process object under a normal working condition, and comprises the following steps:
step (1): collecting 1000 sample data x by using measuring instrument in chemical process object under normal operation state of the chemical process1，x2，…，xN。
Step (2): determining the first 200 sample data x1，x2，…，xnAfter belonging to the same production working condition, according to the formula mu0＝(x1+x2+…+xn) N and the above formula (I) respectively calculate the mean vector mu0And the standard deviation vector sigma0。
And (3): using mean vector mu0Sum standard deviation vector σ0For matrix X ═ X1 x2 … xN]∈Rm×NImplementing a normalization process to obtain a new matrix
And will be
The column vectors of the first n columns form a reference matrix X0∈Rm×nWill be
The column vectors of the middle and rear N-N columns form a test matrix Y0∈Rm×(N-n)。
And (4): identifying N sample data x collected in the step (1) by using the steps (4.1) to (4.7)1，x2，…，xNAnd the identified production conditions are labeled as 1, 2, …, C, in sequence.
In step (4), a monitoring index vector psi is drawn0∈RN×1Fig. 3 shows the variation curve of (a). As can be seen from fig. 3, the multi-condition chemical process comprises C ═ 3 production conditions, and is labeled 1, 2, and 3 in sequence. These N sample data x1，x2，…，xNThe production condition attribution can also be conveniently determined according to the step jump curve shown in fig. 3.
And (5): calculating the mean vector μ ═ of all column vectors in matrix X (X)1+x2+…+xN) N and the standard deviation vector σ:
and carrying out standardization processing on the data matrix X by utilizing mu and sigma to obtain a matrix
In the above formula (c), (x)i-μ)·(xi- μ) represents a vector (x)i- μ) and vector (x)i- μ) at the same position.
And (6): according to the production working condition attribution of each sample data identified in the step (4), the matrix is processed
Divided into C data matrices
And (7): determining a main transformation matrix W epsilon R according to the steps (7.1) to (7.4)m×DAnd a sub-transform matrix
It is worth emphasizing that step (4.3) and step (7.3) both involve the implementation of determining parameters according to descending order of eigenvalues, and the implementation of determining parameters is the same. Taking the characteristics in step (4.3) as an example, the specific implementation process is as follows.
Step (A): set d to 1.
Step (B): according to the formula Δd＝(λd-λd+1)/λdCalculating the characteristic value change rate Deltad。
Step (C): judging whether the condition is satisfied
If yes, returning to the step (B) after d is set to d + 1; if not, setting the parameter D to D-D-1; wherein the threshold value
Is limited to the interval [ 0.050.2 ]]An internal value.
And (8): according to the formula
Calculating a score matrix S corresponding to the kth production conditionkThen, calculate S againkMean vector mu of all column vectors inkAnd S ofkCovariance matrix Λk。
And (9): determining the upper control limit U according to the formulas (c) and (d)limAnd control upper limit Qlim。
Step (10): keeping mean vector sum mu, standard deviation vector sigma, main transformation matrix W and auxiliary transformation matrix
Mean vector mu1，μ2，…，μCCovariance matrix Λ1，Λ2，…，ΛCAnd an upper control limit UlimAnd QlimAnd the real-time calling is carried out when the online process monitoring is implemented.
The off-line modeling stage is completed, and then the on-line dynamic process monitoring stage is started, so that the on-line sampling data of the CSTR chemical process object is required to be utilized in real time. When the on-line state monitoring is implemented, the CSTR just started runs under the normal working condition, and the cooling water valve of the CSTR has viscous fault after a period of time, so that the CSTR cannot be adjusted.
Step (11): sample data x of latest sampling moment of multi-condition chemical process object is acquired on linet∈Rm×1And using the mean vector and mu andvector of standard deviation σ vs xtCarrying out standardization to obtain vector
Step (12): according to the formula respectively
And
calculating a principal score vector stAnd the sub-score vector
Then, according to the above-mentioned formula (v) and formula (sixthly), respectively calculating monitoring index UtAnd a monitoring index Qt。
Step (13): judging whether the condition U is satisfiedt≤UlimAnd Qt≤Qlim(ii) a If yes, the chemical process does not have an abnormal state during operation at the current sampling moment, and the step (11) is returned to continue to monitor the state of the next latest sampling moment; if not, executing the step (14) so as to decide whether an abnormal state occurs;
step (14): returning to the step (11) to continue to carry out state detection on the next latest sampling moment, and if the monitoring indexes of the continuous 3 sampling moments do not meet the judgment condition in the step (13), generating a fault in the production process and triggering a fault alarm; otherwise, an abnormal state alarm is not triggered, and the step (11) is returned to continue to carry out state monitoring on the next latest sampling moment.
As shown in FIG. 4, the method of the present invention is implemented according to the monitoring index U when performing on-line monitoringtAnd QtPlotted in the graph, the horizontal line in FIG. 4 is the control upper limit U corresponding to eachlimAnd Qlim. As can be seen from FIG. 3, the method of the present invention can determine a plurality of production conditions of the chemical process object and the sample data thereof very conveniently. As can be seen from FIG. 4, it is not necessary to inform the production condition to which the sample data belongs to directly utilize U during online monitoringtAnd QtMonitoring is carried out, and after the actuator fault occurs, the actuator fault can be successfully detected and identified.
The above embodiments are merely illustrative of specific implementations of the present invention and are not intended to limit the present invention. Any modification of the present invention within the spirit of the present invention and the scope of the claims will fall within the scope of the present invention.
Claims (2)
1. A chemical process monitoring method integrating multi-working-condition identification and fault detection is characterized by comprising the following steps: firstly, the off-line modeling stage comprises the following steps (1) to (10);
step (1): collecting N sample data x under normal operation state of chemical process by using measuring instrument in chemical process object1，x2，…，xNWherein x isi∈Rm×1Represents the ith sample data, Rm×1A real number vector representing dimension m × 1, m being the total number of measured variables of the chemical process object, with a subscript i ═ 1, 2, …, N;
step (2): determining the first n sample data x1，x2，…，xnAfter belonging to the same production working condition, according to the formula mu0＝(x1+x2+…+xn) N and the following formula (i) calculate the mean vector mu separately0And the standard deviation vector sigma0：
Wherein N is less than N and (x)j-μ0)·(xj-μ()) Represents a vector (x)j-μ0) And vector (x)j-μ0) Where the elements in the same position are multiplied together, with the subscript number j ═ 1, 2, …, n;
and (3): using mean vector mu0Sum standard deviation vector σ0For matrix X ═ X1 x2 … xN]∈Rm×NImplementation criteriaProcessing to obtain new matrix
And will be
The column vectors of the first n columns form a reference matrix X0∈Rm×nWill be
The column vectors of the middle and rear N-N columns form a test matrix Y0∈Rm×(N-n)Wherein R ism×nRepresenting a matrix of real numbers in m x n dimensions, Rm×(N-n)A matrix of real numbers representing dimensions mx (N-N);
and (4): identifying the N sample data x acquired in step (1) by using the following steps (4.1) to (4.7)1，x2，…，xNAnd marking the plurality of identified production conditions as 1, 2, …, and C in sequence, wherein C is the total number of the identified production conditions;
step (4.1): according to the formula
Calculating a symmetric matrix phi epsilon Rm×mWherein A is-1/2And after representing the inverse matrix of the calculation matrix A, carrying out root number opening processing, namely: (A)-1/2)(A-1/2)＝A-1；
Step (4.2): calculating eigenvectors corresponding to all eigenvalues of the symmetric matrix phi, and arranging all eigenvalues in descending order according to magnitude to obtain lambda1≥λ2≥…≥λmCharacteristic value lambda1，λ2，…，λmThe corresponding characteristic vectors are v in sequence1，v2，…，vm；
Step (4.3): according to the characteristic value lambda1，λ2，…，λmDetermining the parameter D0；
Step (4.4): will D0A characteristicVector quantity
Composition feature matrix
Then, the transformation matrix is calculated
Step (4.5): according to the formula
Calculating score matrix S0Then, calculate S again0Covariance matrix of
Step (4.6): according to the formula
Calculating a monitoring index vector psi0∈RN×1Then, drawing psi0In which
diag { } denotes an operation of converting a matrix diagonal element in braces into a column vector;
step (4.7): according to the variation curve drawn in the step (4.6), a plurality of production working conditions are identified and marked as 1, 2, … and C in sequence, and N sample data x collected in the step (1) are determined1，x2，…，xNBelonging to production working conditions;
and (5): calculating a mean vector mu and a standard deviation vector sigma of all column vectors in the matrix X, and normalizing the data matrix X by using the mu and the sigma to obtain a matrix
And (6): according to the production working condition attribution of each sample data in the step (4), the matrix is divided into a plurality of matrixes
Divided into C data matrices
Wherein the content of the first and second substances,
is a matrix under the k production condition,
Represents mxNkReal number matrix of dimension, NkThe total number of sample data, subscript number k, which indicates the kth production regime is 1, 2, …, C;
and (7): determining a main transformation matrix W epsilon R according to the steps (7.1) to (7.4) shown as followsm×DAnd a sub-transform matrix
Step (7.1): calculating the symmetry matrix theta epsilon R according to the formula shown belowm×m：
In the above formula II, matrix
Is composed of a matrix
Removing matrix
After left overThe C-1 matrixes are combined;
step (7.2): calculating eigenvectors corresponding to m eigenvalues in the symmetric matrix theta, and arranging the m eigenvalues in descending order according to the magnitude to obtain eta1≥η2≥…≥ηmCharacteristic value eta1，η2，…，ηmThe corresponding characteristic vectors are w in turn1，w2，…，wm；
Step (7.3): according to the characteristic value eta1，η2，…，ηmDetermining a parameter D according to the change condition;
step (7.4): the first D feature vectors p1，p2，…，pDThe constituent master feature matrix P ═ P1，p2，…，pD]The last m-D eigenvectors pD+1，pD+2，…，pmComposing a sub-feature matrix
Step (7.5): calculating main transformation matrix
And a sub-transform matrix
And (8): according to the formula
Calculating score matrix SkThen, calculate S againkMean vector mu of all column vectors inkAnd S ofkCovariance matrix Λk；
And (9): the upper control limit U is determined according to the formula shown belowlimAnd control upper limit Qlim：
In the above formula, the first and second carbon atoms are,
the value of the chi-square distribution with the degree of freedom D under the condition of the confidence degree alpha is represented,
representing a degree of freedom of
The value of the chi-square distribution under the condition of the confidence degree alpha,
step (10): keeping mean vector sum mu, standard deviation vector sigma, main transformation matrix W and auxiliary transformation matrix
Mean vector mu1，μ2，…，μCCovariance matrix Λ1，Λ2，…，ΛCAnd an upper control limit UlimAnd QlimTo be called in real time when online process monitoring is implemented;
secondly, after the off-line modeling stage is finished, sample data x measured in real time can be utilizedt∈Rm×1Implementing online monitoring on the operating state of the multi-working-condition chemical process, comprising the following steps (11) to (14);
step (11): sample data x of latest sampling moment of multi-condition chemical process object is acquired on linet∈Rm×1And using the mean vector sum mu and the standard deviation vector sigma to xtCarrying out standardization to obtain vector
Wherein the subscript t denotes the latest sampling instant;
step (12): according to the formula respectively
And
calculating a principal score vector stAnd the sub-score vector
Then, according to formula Ak＝(st-μk)TΛk(st-μk) Calculating the Mahalanobis squared distance A1，A2，…，ACWherein k is 1, 2, …, C;
step (13): respectively calculating the monitoring indexes U according to the formula shown in the specificationtAnd a monitoring index Qt：
Ut＝min{A1，A2，…，AC} ⑤
In the above formula, min { A }1，A2，…，ACDenotes A1，A2，…，ACA medium to minimum value;
step (14): judging whether the condition U is satisfiedt≤UlimAnd Qt≤Qlim(ii) a If yes, the chemical process does not have an abnormal state during operation at the current sampling moment, and the step (11) is returned to continue to monitor the state of the next latest sampling moment; if not, executing the step (15) so as to decide whether an abnormal state occurs;
step (15): returning to the step (11) to continue to carry out state monitoring on the next latest sampling moment, and if the monitoring indexes of 3 continuous sampling moments do not meet the judgment condition in the step (14), the production process breaks down and a fault alarm is triggered; otherwise, an abnormal state alarm is not triggered, and the step (11) is returned to continue to carry out state monitoring on the next latest sampling moment.
2. The method for monitoring the chemical process integrating the multi-condition identification and the fault detection according to claim 1, wherein the step (4.3) is performed according to a characteristic value λ1，λ2，…，λmDetermining the parameter D0The specific implementation process is as follows:
step (A): setting d to 1;
step (B): according to the formula Δd＝(λd-λd+1)/λdCalculating the characteristic value change rate Deltad；
Step (C): judging whether the condition is satisfied
If yes, returning to the step (B) after d is set to d + 1; if not, setting the parameter D to D-D-1; wherein the threshold value
Is limited to the interval [ 0.050.2 ]]Internal value taking;
furthermore, step (7.3) is based on the characteristic value η1，η2，…，ηmThe implementation of the variation determining parameter D is the same as that of the above steps (a) to (C).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010570615.3A CN111914888A (en) | 2020-06-13 | 2020-06-13 | Chemical process monitoring method integrating multi-working-condition identification and fault detection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010570615.3A CN111914888A (en) | 2020-06-13 | 2020-06-13 | Chemical process monitoring method integrating multi-working-condition identification and fault detection |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111914888A true CN111914888A (en) | 2020-11-10 |
Family
ID=73226189
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010570615.3A Pending CN111914888A (en) | 2020-06-13 | 2020-06-13 | Chemical process monitoring method integrating multi-working-condition identification and fault detection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111914888A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114708712A (en) * | 2022-03-31 | 2022-07-05 | 杭州电子科技大学 | Information fusion method for fault detection of chemical process reactor |
| CN116206154A (en) * | 2023-02-17 | 2023-06-02 | 北京科技大学 | Fault classification method for sewage treatment process under unusual working condition |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009070071A (en) * | 2007-09-12 | 2009-04-02 | Toshiba Corp | Learning process abnormality diagnostic device and operator's judgement estimation result collecting device |
| US20140067324A1 (en) * | 2012-09-04 | 2014-03-06 | Taiwan Semiconductor Manufacturing Company, Ltd. | Qualitative fault detection and classification system for tool condition monitoring and associated methods |
| CN109062196A (en) * | 2018-10-31 | 2018-12-21 | 东北大学 | The blast furnace process monitoring of integrated PCA-ICA a kind of and method for diagnosing faults |
| CN109407652A (en) * | 2018-12-10 | 2019-03-01 | 中国石油大学(华东) | Multivariable industrial process fault detection method based on major-minor pca model |
| CN110244692A (en) * | 2019-06-20 | 2019-09-17 | 中国石油大学(华东) | Chemical process small fault detection method |
-
2020
- 2020-06-13 CN CN202010570615.3A patent/CN111914888A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009070071A (en) * | 2007-09-12 | 2009-04-02 | Toshiba Corp | Learning process abnormality diagnostic device and operator's judgement estimation result collecting device |
| US20140067324A1 (en) * | 2012-09-04 | 2014-03-06 | Taiwan Semiconductor Manufacturing Company, Ltd. | Qualitative fault detection and classification system for tool condition monitoring and associated methods |
| CN109062196A (en) * | 2018-10-31 | 2018-12-21 | 东北大学 | The blast furnace process monitoring of integrated PCA-ICA a kind of and method for diagnosing faults |
| CN109407652A (en) * | 2018-12-10 | 2019-03-01 | 中国石油大学(华东) | Multivariable industrial process fault detection method based on major-minor pca model |
| CN110244692A (en) * | 2019-06-20 | 2019-09-17 | 中国石油大学(华东) | Chemical process small fault detection method |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114708712A (en) * | 2022-03-31 | 2022-07-05 | 杭州电子科技大学 | Information fusion method for fault detection of chemical process reactor |
| CN116206154A (en) * | 2023-02-17 | 2023-06-02 | 北京科技大学 | Fault classification method for sewage treatment process under unusual working condition |
| CN116206154B (en) * | 2023-02-17 | 2024-02-13 | 北京科技大学 | Fault classification method for sewage treatment process under unusual working condition |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112527788A (en) | Method and device for detecting and cleaning abnormal value of transformer monitoring data | |
| CN106647718B (en) | Nonlinear industrial processes fault detection method based on the slow signature analysis of Bayes's core | |
| CN108803520A (en) | A kind of dynamic process monitoring method rejected based on the non-linear autocorrelation of variable | |
| CN109491358B (en) | Control performance monitoring method for boiler dynamic information of million-kilowatt ultra-supercritical unit | |
| CN112036089A (en) | Coal mill fault early warning method based on DPC-MND and multivariate state estimation | |
| CN111914888A (en) | Chemical process monitoring method integrating multi-working-condition identification and fault detection | |
| CN109407652A (en) | Multivariable industrial process fault detection method based on major-minor pca model | |
| CN110262450B (en) | Fault prediction method for cooperative analysis of multiple fault characteristics of steam turbine | |
| CN106940808A (en) | A kind of fault detection method based on modified Principal Component Analysis Model | |
| CN105955214B (en) | Batch process fault detection method based on sample time-series and neighbour's affinity information | |
| CN109538311B (en) | Real-time monitoring method for control performance of steam turbine in high-end power generation equipment | |
| CN108960309B (en) | Dynamic process monitoring method based on RBF neural network autocorrelation elimination | |
| CN111914887B (en) | Novel multi-mode chemical process abnormal state detection method | |
| CN108181893B (en) | PCA-KDR-based fault detection method | |
| CN110032799A (en) | A kind of the angle similarity divided stages and monitoring method of microbiological pharmacy process | |
| Deng et al. | Sparse stacked autoencoder network for complex system monitoring with industrial applications | |
| CN110942258B (en) | Performance-driven industrial process anomaly monitoring method | |
| CN103926919A (en) | Industrial process fault detection method based on wavelet transform and Lasso function | |
| CN111913447B (en) | Online data-driven nonlinear chemical process monitoring method | |
| Li et al. | Multiple sensor fault diagnosis for dynamic processes | |
| CN110362063B (en) | Fault detection method and system based on global maintenance unsupervised kernel extreme learning machine | |
| CN111914886B (en) | Nonlinear chemical process monitoring method based on online brief kernel learning | |
| CN109283912B (en) | Distributed dynamic and static cooperative comprehensive monitoring method for pulverizing system of large coal-fired generator set of intelligent power plant | |
| CN114841621A (en) | Method for detecting abnormal operation state of synthetic ammonia process based on online neighbor embedding index | |
| Xinyi et al. | Fault diagnosis in chemical process based on self-organizing map integrated with fisher discriminant analysis |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20240205 Address after: 518000 1002, Building A, Zhiyun Industrial Park, No. 13, Huaxing Road, Henglang Community, Longhua District, Shenzhen, Guangdong Province Applicant after: Shenzhen Wanzhida Technology Co.,Ltd. Country or region after: China Address before: 315211, Fenghua Road, Jiangbei District, Zhejiang, Ningbo 818 Applicant before: Ningbo University Country or region before: China |