CN116068974A

CN116068974A - Distributed process monitoring method and device for inter-block collaborative modeling in block

Info

Publication number: CN116068974A
Application number: CN202310058800.8A
Authority: CN
Inventors: 钟伟民; 彭鑫; 李智; 杜文莉; 钱锋; 曹志兴
Original assignee: East China University of Science and Technology
Current assignee: East China University of Science and Technology
Priority date: 2023-01-18
Filing date: 2023-01-18
Publication date: 2023-05-05

Abstract

The invention relates to the technical field of industrial process monitoring, in particular to a distributed process monitoring method and device for inter-block collaborative modeling in blocks. The method comprises the following steps: the offline modeling phase includes the steps of: decomposing the process data to obtain each sub-block; modeling each sub-block by using a slow feature analysis method in the sub-block; performing typical correlation analysis modeling among sub-blocks; the online monitoring stage comprises the following steps: partitioning new sampled data; substituting the segmented sampling data into a slow characteristic analysis method model in the sub-block and a typical correlation analysis model among the sub-blocks respectively to obtain corresponding characteristic components; and (3) adopting Bayesian reasoning fusion to obtain a comprehensive index as a final monitoring statistic, and comparing the comprehensive index with a corresponding control limit to detect the state of the current sub-block. The invention can detect faults from different characteristic angles of data, and provides a preliminary range of fault positioning so as to improve the detection performance and the fault detection rate of the process.

Description

Distributed process monitoring method and device for inter-block collaborative modeling in block

Technical Field

The invention relates to the technical field of industrial process monitoring, in particular to a distributed process monitoring method and device for inter-block collaborative modeling in blocks.

Background

In the modern industry, ensuring the safe progress of process production and the quality of products and improving the production efficiency are the directions of research for a long time. From the most traditional model-based methods, the actual system production state is supervised for deviations depending on the exact mathematical model of the established process (such as the laws of physics and chemical reaction processes). With the increasing complexity of process systems, it becomes more difficult and costly to build feature models.

People form knowledge-based methods to monitor the process by accumulating knowledge and expert experience about the industrial process. The monitoring results of knowledge-based methods are more intuitive and understandable, but to date, in some process areas, the time it takes for expert experience and knowledge to accumulate is shorter, and the knowledge base is smaller, so knowledge-based methods are difficult to apply.

Unlike the long-term accumulation of knowledge, with the development of modern industrial systems, a large amount of data containing industrial process information is available, and thus research into data-driven process monitoring methods is being conducted well.

At the heart of data-driven process monitoring is mining of data features. In practice, however, different industrial processes, the process data of which exhibit different characteristics or have a plurality of characteristics. The data of the common process industry are mainly characterized as follows:

gao Weixing. The high dimensionality of the data is unavoidable and the trend in modern industrial processes is toward large scale and

complex developments. The whole production flow consists of various components and operation devices, and in each part, important variables which can be measured exist, so that the dimension of data collected by the process is higher and higher.

A non-gaussian distribution. The use of classical process monitoring methods is mostly based on the premise of gaussian processes. However, in actual production, not all variables are subject to a gaussian distribution, and some may be subject to different types of non-gaussian distributions due to non-gaussian noise or feedback control loops, etc. Therefore, many corresponding methods of monitoring non-gaussian process data have also been developed.

Non-linearities. The presence of non-linearities is common, and in a process, there is a linear correlation between some process variables, and in some cases not, there may be a non-linear correlation between process and quality variables. Various nonlinear relations exist in the process data, analysis is carried out according to specific characteristics, and a corresponding process monitoring model is established.

Time-varying properties. In industrial processes, the operating conditions are affected in many ways, such as fluctuations in the feed composition, ageing of the process equipment, scheduling of production plans, etc., which change over time. For this purpose, corresponding solutions, such as process monitoring schemes for adaptive techniques, multi-modal methods, etc., are also proposed.

And (3) autocorrelation. In addition to the correlation between variables, there is also a dynamic nature of the variables themselves, with the current measured values of the variables being auto-correlated with their past values. The time series correlation of variables may be related to control loops, feedback loops, process noise, etc., in addition to the nature of the process. Some faults affect the dynamics of the process, so mining of data dynamics is also an important aspect of process monitoring.

Today, industrial systems are increasingly complex, data is increasingly collected, and features of the data are increasingly complex. How to better mine the data features is a key to improving the process monitoring performance.

For a large-scale process, the distributed typical correlation analysis (CCA) process monitoring based on partial sub-block communication is carried out, process decomposition is carried out by combining with an actual process background, a topology matrix for describing the connection between sub-blocks is provided, CCA modeling is carried out on the sub-blocks with connection and information interaction according to the topology matrix, the correlation between process data of the two sub-blocks is excavated, and the process monitoring is realized by observing the space change constructed based on residual vector.

However, the distributed CCA process monitoring method based on partial sub-block communication only focuses on the mining of the correlation between sub-blocks, and the measurement data of each sub-block also contains a lot of process information in the current range, so that the local data can be extracted to construct a principal component feature space for supervision.

Moreover, distributed CCA modeling based on partial sub-block communication may face a serious problem, when a certain sub-block is topologically communicated with only a certain sub-block, if a problem occurs in data transmission, such as line interruption, etc., the data of the current sub-block is not used for process monitoring.

It is therefore desirable to propose a solution to the above-mentioned problems.

Disclosure of Invention

The invention aims to provide a distributed process monitoring method and device for intra-block inter-block collaborative modeling, which solve the problem of poor monitoring performance of a distributed CCA process in the prior art.

In order to achieve the above object, the present invention provides a distributed process monitoring method for inter-block collaborative modeling in a block, which includes two stages of offline modeling and online monitoring, wherein the offline modeling stage specifically includes the following steps:

step S1: decomposing the process data to obtain each sub-block, and constructing a topology matrix C;

Step S2: modeling each sub-block by a slow feature analysis method in the sub-block, extracting slow features and reserving a projection matrix;

step S3: respectively calculating statistics according to the extracted slow features, and obtaining corresponding control limits through kernel density estimation;

step S4: carrying out typical correlation analysis modeling among the sub-blocks according to the topology matrix C, extracting typical correlation components and reserving a projection matrix;

step S5: generating residual vectors according to typical relevant components, calculating statistics, and obtaining corresponding control limits through kernel density estimation;

the online monitoring stage specifically comprises the following steps:

step S6: carrying out corresponding block division on the new sampling data according to the step S1, and decomposing to obtain each sub-block;

step S7: substituting the segmented sampling data into a slow characteristic analysis method model in the sub-block and a typical correlation analysis model among the sub-blocks respectively to obtain corresponding characteristic components;

step S8: respectively calculating statistics of models corresponding to the slow feature analysis method model and the typical correlation analysis model;

step S9: for different detection results of multiple models of the same sub-block, a Bayesian inference fusion is adopted to obtain a comprehensive index ET ² As final monitoring statistic and comparing with corresponding control limit to detect current sub-block state;

step S10: and calculating a fault detection index FD of the whole process, comparing the fault detection index FD with a corresponding control limit, and when the fault detection index FD exceeds the control limit, indicating that a fault occurs in the process, otherwise, considering the running state of the process as a normal state.

In one embodiment, the step S1 further includes:

and decomposing the process data by combining the process background and the mechanism knowledge to obtain each sub-block, wherein the corresponding expression is as follows:

X＝[X ₁ ,X ₂ ,…,X _i …,,X _B ]

wherein ,X_i The ith sub-block, and B is the number of sub-blocks.

In an embodiment, in the step S2, the slow feature extracted by each sub-block corresponds to the following expression:

s _i ＝W _i X _i

wherein ,s_i Slow feature for the ith sub-block, W _i For the weight matrix corresponding to the ith sub-block, X _i Is the i-th sub-block.

In one embodiment, the statistics in step S3 correspond to the following expressions:

wherein ,

statistics corresponding to the ith sub-block for monitoring the change condition of components representing the essence of the process state in the subspace, s _i Is a slow feature of the I-th sub-block.

In an embodiment, the kernel density estimation

The corresponding expression is:

Wherein y is the data to be estimated, y _i For the observations of the process data, n is the number of samples, h is the smoothing parameter, and K (·) is the kernel function.

In an embodiment, in the step S4, a representative correlation analysis modeling is performed between the sub-blocks and the sub-blocks having information interaction with the sub-blocks according to the topology matrix C, and the obtained representative correlation component has the following expression:

wherein ,A_ij and B_ij Projection matrix for sub-block i and sub-block j, u _ij and v_ij Typical components of sub-block i and sub-block j, respectively.

In one embodiment, in the step S5, a residual vector is created based on the extracted typical correlation component, and the corresponding expression is:

wherein ,r_ij For the residual vector between sub-block i and sub-block j, Λ _k,ij A typical correlation coefficient diagonal matrix, X, for the top k pairs of correlation variables of sub-block i and sub-block j _i For the ith sub-block, X _j Is the j-th sub-block;

in the step S5, statistics are constructed based on residual vectors, and the corresponding expression is:

wherein ,

for statistics constructed based on residual vectors between sub-block I and sub-block j, I _k Is a K-order unit array.

In one embodiment, the final monitoring statistic ET in step S9 ² The corresponding expression is:

wherein ,

final monitoring statistic for ith sub-block, x _new For online data +.>

Conditional probability for an abnormal operating state of the process, +.>

For the probability of failure occurrence, P (c _ij )＝c _ij Is a coefficient representing the connection relationship between sub-blocks based on the topology matrix C.

In one embodiment, in the step S9, the control limit CL corresponding to the final monitoring statistic of the ith sub-block _i The prior probability (1-alpha) of abnormal working conditions of the process is given.

In one embodiment, in the step S10, the fault detection index FD of the whole process is expressed as follows:

wherein the control limit of the failure detection index FD is 1.

To achieve the above object, the present invention provides a distributed process monitoring apparatus for intra-block inter-block collaborative modeling, including:

a memory for storing instructions executable by the processor;

a processor for executing the instructions to implement the method as claimed in any one of the preceding claims.

To achieve the above object, the present invention provides a computer readable medium having stored thereon computer instructions, wherein the computer instructions, when executed by a processor, perform a method as set forth in any of the above.

The distributed process monitoring method and device for collaborative modeling among blocks in the block can provide process monitoring, detect faults from different characteristic angles of data, and provide a preliminary range of fault positioning so as to improve the detection performance and the fault detection rate of the process.

Drawings

The above and other features, properties and advantages of the present invention will become more apparent from the following description of embodiments taken in conjunction with the accompanying drawings in which like reference characters designate like features throughout the drawings, and in which:

FIG. 1 discloses a flow chart of a distributed process monitoring method for intra-block inter-block collaborative modeling in accordance with an embodiment of the present invention;

FIG. 2 discloses a schematic diagram of a distributed process monitoring method for intra-block inter-block collaborative modeling in accordance with an embodiment of the present invention;

FIG. 3 discloses a reference simulation model flow diagram No. 1 in accordance with an embodiment of the present invention;

FIG. 4 discloses a schematic diagram of fault detection results of a fault 3 of a BMS1 process based on an SFA method;

fig. 5 discloses a schematic diagram of a fault detection result of a distributed CCA process monitoring method based on partial sub-block communication on a fault 3 of a BMS1 process;

fig. 6 discloses a schematic diagram of a fault detection result of a fault 3 of a BMS1 process according to an embodiment of the present invention;

FIG. 7 is a schematic diagram of the results of fault detection of a sub-block 1 to a TE process fault 3 in a distributed CCA process monitoring method based on partial sub-block communication;

FIG. 8 is a schematic diagram of fault detection results of a sub-block 1 to a fault 16 of a TE process according to an embodiment of the present invention;

FIG. 9 is a schematic diagram of fault detection results of a sub-block 2 to a TE process fault 3 in a distributed CCA process monitoring method based on partial sub-block communication;

FIG. 10 is a diagram showing the fault detection results of sub-block 2 to the fault 16 of TE process according to one embodiment of the present invention;

FIG. 11 is a schematic diagram of fault detection results of a sub-block 3 to a TE process fault 3 in a distributed CCA process monitoring method based on partial sub-block communication;

FIG. 12 discloses a schematic diagram of fault detection results of a sub-block 3 on a fault 16 of a TE process according to an embodiment of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

So that those skilled in the art can appreciate the features and effects of the present invention, a general description and definition of the terms and expressions set forth in the specification and claims follows. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs, and in the event of a conflict, the present specification shall control.

In this context, not all possible combinations of the individual technical features in the individual embodiments or examples are described in order to simplify the description. Accordingly, as long as there is no contradiction between the combinations of these technical features, any combination of the technical features in the respective embodiments or examples is possible, and all possible combinations should be considered as being within the scope of the present specification.

Fig. 1 discloses a flowchart of a distributed process monitoring method for intra-block inter-block collaborative modeling according to an embodiment of the present invention, fig. 2 discloses a schematic diagram of a distributed process monitoring method for intra-block inter-block collaborative modeling according to an embodiment of the present invention, and as shown in fig. 1 and fig. 2, the present invention provides a distributed process monitoring method for intra-block inter-block collaborative modeling, which includes two stages of offline modeling and online monitoring, and the offline modeling stage specifically includes the following steps:

step S2: modeling a slow feature analysis method (SFA) in each sub-block, extracting slow features and reserving a projection matrix;

Step S4: modeling a typical correlation analysis (CCA) among the connected sub-blocks according to the topology matrix C, extracting typical correlation components and reserving a projection matrix;

the online monitoring stage specifically comprises the following steps:

step S7: substituting the segmented sampling data into a slow feature analysis method (SFA) model in the sub-block and a typical correlation analysis (CCA) model among the sub-blocks respectively to obtain corresponding feature components;

The invention provides a distributed process monitoring method of inter-block collaborative modeling in a block, namely a distributed process monitoring method (PDCCA-SFA) of inter-block collaborative modeling in a block based on SFA and CCA, which is based on distributed CCA process monitoring based on partial sub-block communication, and combines the SFA method to build an SFA local model in each sub-block, namely to build a local model by using data stored by local measurement for each sub-block to supervise, and build a CCA model among the sub-blocks of information interaction, faults are detected from different characteristic angles of the data, so that the detection performance of the process is improved, the SFA method (slow feature analysis method) is different from the existing principal component analysis technology (PCA) method, the SFA extracts the component which changes most slowly in the data representing the essence of the process, and simultaneously considers the process dynamics, and as the SFA algorithm explores the dynamic change of the process from the internal angle of the data, the extracted slow features not only comprise steady state information of the process, but also comprise dynamic information of the process and can be used for the dynamic modeling process of abnormal detection.

The SFA method comprises the following specific principles:

let p-dimensional input signal x (t) = (x) ₁ (t),x ₂ (t),…,x _p (t)) ^T The SFA method hopes to find a transformation function f (x) = (f) ₁ (x),f ₂ (x),…,f _p (x)) ^T The input signal can be converted into a series of output signals s (t) =f (x (t)) that vary as slowly as possible, with the optimization objective being:

wherein ,

representing the output signal s _j The first derivative of (t) is defined by +.>

The calculation result shows that the method comprises the steps of,<·> _t represent time average, deltas _j (t) represents the output signal s _j The rate of change of (t),<s _j (t)> _t＝0 and

Constraints representing zero mean and unit variance, respectively, to avoid getting trivial solutions

To ensure that the output signals are uncorrelated and output in a sequence, i.e. the first output signal varies most slowly and the second output signal varies second slowly.

The SFA method is a linear method that assumes that there is a correlation between the input and output variables, and thus the slow features s are extracted _j (t) can be regarded as a linear combination of a series of input signals

The expression for the slow features thus extracted is:

s(t)＝Wx(t)

wherein W= [ W ₁ ,w ₂ ,…,w _p ] ^T Is a weight matrix.

Given the input vector x, the whitening process is required to be performed before the input vector x is used for the slow feature analysis algorithm, so that the whitening process is eliminatedCorrelation between variables, whitening processing by singular value decomposition (Singular Value Decomposition, SVD), b=<xx ^T > _t Representing the expectation of the covariance matrix of the input vector x, which is subjected to singular value decomposition:

B＝UΛU ^T

Wherein Λ is a diagonal matrix formed by eigenvalues, and U is a matrix formed by corresponding eigenvectors, so that whitening data z can be obtained:

z＝Λ ^-1/2 U ^T x＝Qx

satisfying cov (z) =<zz ^T >＝I _p ，I _p As a unitary matrix, q=Λ ^-1/2 U ^T Is a whitening matrix.

Thus, the optimization objective of the SFA algorithm can be reduced to finding a matrix P that satisfies the following transformations:

s＝Pz

at the same time satisfy

ss ^T ＝I _p

Push out<PP ^T >＝I _p The matrix P is an orthogonal matrix.

Accordingly, to extract slow features, singular value decomposition is performed on the covariance matrix of the first derivative of whitened data z:

thus, the expressions for the weight matrix W and slow feature s can be obtained as:

W＝PQ

s＝Wx。

the above steps of the present invention will be described in detail with reference to fig. 1 and 2. It is understood that within the scope of the present invention, the above-described technical features of the present invention and technical features specifically described below (e.g., in the examples) may be combined with each other and associated with each other, thereby constituting a preferred technical solution.

Step S1: and decomposing the process data to obtain each sub-block, and constructing a topology matrix C.

The process is decomposed by combining the process background and the mechanism knowledge, and each obtained sub-block corresponds to the expression:

X＝[X ₁ ,X ₂ ,…,X _i …,,X _B B

wherein ,X_i The ith sub-block, and B is the number of sub-blocks.

The method for dividing the blocks based on the process knowledge has different dividing difficulties of different production processes, has certain feasibility, and is also beneficial to the primary range positioning of the faults after the faults are detected.

After dividing the process into B sub-blocks according to mechanism knowledge and expert experience, defining a corresponding process topology matrix C, wherein the topology matrix C is provided for describing the connection relation among the sub-blocks, and considering that different process flows have various structures in the actual industrial process, not all the sub-blocks can mutually transmit information, and not all the sub-blocks have internal relations such as hierarchical progression in structure.

Therefore, compared with distributed CCA process monitoring of all sub-blocks in a pairwise interaction mode, the invention provides a topology matrix C for describing a connection mode among the sub-blocks, and models part of the sub-blocks in communication based on the topology matrix C, so that the communication load of data transmission is further reduced, and meanwhile, the established distributed CCA process monitoring is more in line with the actual industrial process.

Each element in the topology matrix C represents whether there is information interaction between the two corresponding sub-blocks, if so, it is set to 1, otherwise, it is set to 0. Since an element represents a relationship between two sub-blocks, the elements on the diagonal are all set to 0.

Step S2: and modeling a slow feature analysis method in each sub-block, extracting slow features and reserving a projection matrix.

The expression of the slow feature extracted by each sub-block is:

s _i ＝W _i X _i

wherein ,s_i Slow feature for the ith sub-block, W _i And the weight matrix corresponding to the ith sub-block.

Step S3: and respectively calculating statistics according to the extracted slow features, and acquiring corresponding control limits through kernel density estimation.

The expression of the statistics calculated from the extracted slow features is:

wherein ,

the statistics corresponding to the ith sub-block are used for supervising the change condition of the components representing the process state essence in the subspace, and the corresponding control limit is obtained by the kernel density estimation.

Based on statistics constructed by slow features, the change condition of components representing the process state essence in subspaces can be monitored, so that fault detection is realized, and the corresponding control limit can be obtained by nuclear density estimation.

Nuclear density estimation (Kernel Density Estimation, KDE) is often used to estimate the distribution characteristics of data as a non-parametric method

The mathematical description is:

wherein ,

for kernel density estimation, y is the data to be estimated, y _i For the observations of the process data, n is the number of samples, h is the smoothing parameter, and K (·) is the kernel function.

There are many kinds of kernel functions, and the commonly used kernel functions are gaussian functions:

step S4: and carrying out sub-block-to-sub-block typical correlation analysis modeling on the sub-blocks with connections according to the topology matrix C, and extracting typical correlation components and preserving a projection matrix.

CCA modeling is performed on the sub-blocks and the sub-blocks with information interaction with the sub-blocks according to the topology matrix C, and the obtained expression of typical components is as follows:

Step S5: and generating residual vectors according to the typical correlation components, calculating statistics, and obtaining corresponding control limits through kernel density estimation.

Residual vectors established based on the extracted typical components correspond to the following expressions:

wherein ,r_ij For the residual vector between sub-block i and sub-block j, Λ _k,ij A typical correlation coefficient diagonal matrix for the top k pairs of correlation variables for sub-block i and sub-block j;

statistics constructed based on residual vectors, the corresponding expressions are:

wherein ,

The statistics constructed based on the residual vector is used for supervising the change condition of the characteristics reflecting the relationship among the sub-blocks in the subspace, so that fault detection is realized, and the corresponding control limit can be obtained by the kernel density estimation.

The calculation manners corresponding to the step S6 to the step S8 in the online monitoring stage are the same as those in the offline modeling stage, and are not described here again.

Step S9: for different detection results of multiple models of the same sub-block, a Bayesian inference fusion is adopted to obtain a comprehensive index ET ² As a final monitoring statistic and compared with a corresponding control limit to detect the state of the current sub-block.

For the same sampling point, the SFA model and the corresponding CCA model of the current sub-block may have different detection results, and Bayesian reasoning is selected to fuse the detection results of the current sub-block to obtain the final monitoring statistic ET of each sub-block ² 。

Final monitoring statistic ET ² The corresponding expression is:

wherein ,

the final monitoring statistic (i.e., the composite index) for the ith sub-block, x _new For online data (i.e. new query samples), -, etc>

Conditional probability for an abnormal operating state of the process, +.>

For the probability of failure occurrence, P (c _ij )＝c _ij Is a coefficient representing a connection relationship between sub-blocks based on the topology matrix C;

conditional probability

The expression of (2) is:

wherein ,

statistics for the ith sub-block->

A corresponding control limit;

probability of failure occurrence

The expression of (2) is:

wherein ,

for the a priori probability of an abnormal operating condition,

For the probability that the on-line data corresponds,

is the prior probability of the normal operation state.

Control limit CL corresponding to final monitoring statistics of the ith sub-block _i The final monitoring statistics of the ith sub-block for the prior probability (1-alpha) of the process occurrence of abnormal conditions

If the control time limit is higher than the corresponding control time limit, judging that the subblock is abnormal, otherwise, considering that the subblock is operated under the normal working condition.

Since the whole production process is affected no matter which sub-block fails from the viewpoint of monitoring the whole process, the whole process should be alerted from the global viewpoint, and therefore, an overall index is designed to represent the monitoring condition of the whole process, and the expression of the fault detection index FD of the whole process is as follows:

wherein, according to the definition of the fault detection index FD, the control limit is 1.

The invention will now be described in detail with reference to the drawings and specific examples. The present embodiment is implemented on the premise of the technical scheme of the present invention, and a detailed implementation manner and a specific operation process are given, but the protection scope of the present invention is not limited to the following examples.

In the embodiment, the distributed process monitoring method based on the intra-block collaborative modeling of the SFA and the CCA is applied to the urban sewage treatment process and is compared and analyzed with the SFA and the distributed CCA method based on partial sub-block communication.

The urban sewage treatment process is a general reference simulation model No.1 (Benchmark Simulation Model No.1, BSM 1), and the BSM1 reference simulation model is a sewage treatment process model proposed by European Union scientific and technical cooperation organization and is widely applied to the sewage treatment process research of personnel and institutions around the world.

In practice, the BSM1 model consists mainly of two models, an activated sludge process model and a double-exponential settling model, respectively. The BSM1 reference simulation model mainly takes carbon and nitrogen in the wastewater to be removed, so that the treated wastewater can reach the secondary discharge standard.

Fig. 3 shows a flow chart of a reference simulation model No.1 according to an embodiment of the present invention, and as shown in fig. 3, the flow structure of the BSM1 is composed of a bioreactor and a secondary sedimentation tank. The bioreactor consists of 5 uniformly mixed units, wherein the first 2 units are anaerobic closed containers (called anoxic tanks or anaerobic tanks), and the last 3 units are aerobic exposure containers (called aeration tanks or aerobic tanks).

In this model, the organic matter is mainly consumed by the vital movements of the microorganisms, while the nitrogen is removed by nitrification by the microorganisms in the aerobic zone and denitrification in the anaerobic zone.

The total height of the secondary sedimentation tank is 4 meters, the secondary sedimentation tank is divided into 10 layers with equal height, and no chemical reaction occurs in the secondary sedimentation tank, and the secondary sedimentation tank mainly has physical filtration effect.

And water in the reaction unit flows out of the fifth reaction tank, one part flows in from the sixth layer of the secondary sedimentation tank, and the other part flows back to the inlet of the reaction unit through a pipeline to complete internal circulation. In the secondary sedimentation tank, the wastewater meeting the sewage discharge standard after sedimentation is discharged to the river from the top of the secondary sedimentation tank, and part of the sludge deposited from the bottom is sent back to the first reaction tank for sludge circulation, and the other part is discharged to landfill.

The BSM1 model introduced three failures, as shown in table 1, the first failure was a step change in the maximum growth rate of autotrophic bacteria of the first aerobic reaction tank, from 0.5 down to 0.3. Because the life activities of autotrophic bacteria are weakened, complex biochemical reactions in the urban sewage treatment process can be influenced, and the state of the whole system is changed.

The second failure is that the oxygen transfer coefficient of the biological reaction tank 4 is reduced to half of the original one. The oxygen transfer coefficient has a great relation with the concentration of dissolved oxygen in the reaction tank, and the concentration of the dissolved oxygen is an important element of the reaction in the urban sewage treatment process, so that the sewage treatment effect is related. Because the microbial activity for removing organic matters and the denitrification of heterotrophic bacteria are not separated from the participation of oxygen.

The third failure is that the sno_2 sensor of the bio-reaction tank 2 is shifted by +0.5. The SNO_2 sensor is used for measuring the concentration of nitric acid and nitrous acid in the biological reaction tank 2, the measured value is input into the controller, and the controller adjusts the internal reflux according to the deviation, so that the system operates normally. Therefore, if the measured value is inaccurate, the purification effect of water is affected, resulting in the waste water failing to meet the discharge standard.

TABLE 1 three failures of BSM1 procedure

Failure of	Description of the invention	Type(s)
			1	Maximum growth rate of autotrophic bacteria of the first aerobic reaction tank	Step
2	Oxygen transfer coefficient of biological reaction tank 4	Step
			3	SNO_2 measurement of biological reaction tank 2	Step

Firstly, according to the sewage treatment process and structure of a BSM1 model, 15 process measurement data for monitoring are divided into 3 sub-blocks, wherein the sub-block 1 is a relevant water quality measurement variable of treated effluent, the sub-block 2 is a water quality component measurement variable of sewage inflow and reaction tank 3, the sub-block 3 is a measurement variable of reaction tanks 4 and 5, and the measurement variable and the sub-block division are respectively shown in tables 2 and 3:

table 2 measurement variables of BSM1 procedure

TABLE 3 sub-block partitioning of BSM1 process variables

After dividing the process data into sub-blocks, determining the expression of the topology matrix C according to the process knowledge as follows:

According to the urban sewage treatment process, the inflow water respectively flows through two anaerobic tanks and three aerobic tanks, organic matters and nitrogen are removed, the inflow water enters a secondary sedimentation tank, and the clarified outflow water is discharged to a river, so that the topological coefficients of a sub-block 2 representing inflow water and a third reaction tank and a sub-block 3 representing a fourth reaction tank and a fifth reaction tank are set to be 1, and the topological coefficients of the sub-block 3 to the sub-block 1 representing the water quality component of the outflow water are also set to be 1, so that in the embodiment, 2 CCA models are established.

The number of typical relevant components is set to be 3, and the prior probability alpha of the normal operation state of the process is 99%.

As shown in table 4, the detection results of three faults of the BSM1 process based on SFA process monitoring, distributed CCA process monitoring (PDCCA) of partial sub-block communication and distributed process monitoring (PDCCA-SFA) of inter-block collaborative modeling are shown, and as can be seen from table 4, for the second fault, the oxygen transfer coefficient of the 4 th biological reaction tank is reduced to half of the original, and the three methods can detect anomalies;

for the first fault, the maximum growth rate of autotrophic bacteria of the first aerobic reaction tank is changed in a step manner, the fault can be detected based on an SFA (small form factor) method, the detection effect of a CCA (clear solution analysis) method based on partial sub-block connection and a proposed method for collaborative modeling between blocks in a block is generally only 60% and 65%, and the analysis reason is probably that the fault has a wider influence range, so that a global SFA model can have a good detection effect, and the detection effect of a local model based on the block is poor;

For the third fault, the SNO_2 sensor of the biological reaction tank 2 is offset, the monitoring method based on SFA process monitoring and the monitoring method based on inter-block collaborative modeling have good detection results, the fault detection rate is 96% and 95% respectively, and the distributed CCA monitoring method based on partial sub-block communication can hardly detect the fault, so that the SFA can effectively detect the fault.

TABLE 4 BSM1 procedure failure detection Rate for SFA, PDCCA and PDCCA-SFA

Taking the third fault as an example, further analysis:

the third failure is specifically a 0.5 offset of the sno_2 sensor of biological reaction tank 2.

Fig. 4 to 6 respectively disclose the fault detection results of the three process detection methods for the fault 3 of the BMS1 process, and as shown in fig. 4 and 6, both the SFA-based monitoring method and the intra-block collaborative modeling-based monitoring method (PDCCA-SFA) of the present invention can detect the fault and issue an alarm after the fault occurs. As shown in fig. 5, the distributed CCA process monitoring method (PDCCA) based on partial block communication does not substantially detect a failure.

Comparing the fault detection results of the sub-blocks of the PDCCA and the PDCCA-SFA on the fault 3 of the TE process, fig. 7, 9 and 11 respectively disclose the detection results of the sub-blocks 1-3 of the PDCCA, and fig. 8, 10 and 12 respectively disclose the detection results of the sub-blocks 1-3 of the monitoring method (PDCCA-SFA) based on inter-block collaborative modeling in blocks, so that three sub-blocks of the distributed CCA monitoring method based on partial sub-block communication can not detect faults, and the sub-block 1 of the distributed process monitoring method based on inter-block collaborative modeling in blocks can detect faults, which means that the local SFA model is built for each sub-block to improve the monitoring performance of the process.

Because the third fault is that the sensor for measuring the concentration of the nitric acid and the nitrous acid is deviated, the internal reflux regulated by the controller is changed, so that the sewage treatment effect is affected, and the fault can be detected by the local SFA model of the sub-block 1 of which the monitoring variable is the water quality component of the effluent.

According to the distributed process monitoring method and device for inter-block collaborative modeling, after process decomposition is considered, the local measurement data of each sub-block can be used for establishing a CCA model among the sub-blocks, rich process information contained in the local measurement data can be mined through slow feature analysis and used for detecting faults in the current range, so that after the process decomposition, a local SFA model is established for each sub-block to detect faults of a specific unit except for establishing the CCA model among the sub-blocks according to a topology matrix, then detection results of the SFA model in the sub-block and the CCA model among the sub-blocks of each sub-block are integrated through Bayesian reasoning, process monitoring is achieved, and effectiveness of the method provided by the invention is verified through a BSM1 process simulation experiment.

The invention provides a distributed process monitoring device for intra-block inter-block collaborative modeling. A distributed process monitoring device co-modeled among blocks within a block may include an internal communication bus, a processor, a Read Only Memory (ROM), a Random Access Memory (RAM), a communication port, and a hard disk. The internal communication bus may enable data communication between the distributed process monitoring device components of the intra-block co-modeling. The processor may make the determination and issue the prompt. In some embodiments, a processor may be comprised of one or more processors.

The communication ports enable data transmission and communication between the intra-block co-modeled distributed process monitoring devices and external input/output devices. In some embodiments, the intra-block inter-block co-modeled distributed process monitoring device may send and receive information and data from a network through a communication port. In some embodiments, the intra-block co-modeled distributed process monitoring device may communicate and transfer data between the input/output terminals and external input/output devices in a wired fashion.

The intra-block co-modeled distributed process monitoring device may also include different forms of program storage units and data storage units, such as hard disks, read Only Memories (ROM) and Random Access Memories (RAM), capable of storing various data files for computer processing and/or communication, and possible program instructions for execution by the processor. The processor executes these instructions to implement the main part of the method. The result processed by the processor is transmitted to an external output device through a communication port and is displayed on a user interface of the output device.

For example, the implementation process files of the distributed process monitoring device for intra-block inter-block collaborative modeling described above may be computer programs, stored on a hard disk, and recorded into a processor for execution to implement the methods of the present invention.

When the process file of the distributed process monitoring method for intra-block inter-block collaborative modeling is a computer program, the process file may also be stored in a computer readable storage medium as an article of manufacture. For example, computer-readable storage media may include, but are not limited to, magnetic storage devices (e.g., hard disk, floppy disk, magnetic strips), optical disks (e.g., compact Disk (CD), digital Versatile Disk (DVD)), smart cards, and flash memory devices (e.g., electrically erasable programmable read-only memory (EPROM), cards, sticks, key drives). Moreover, various storage media described herein can represent one or more devices and/or other machine-readable media for storing information. The term "machine-readable medium" can include, without being limited to, wireless channels and various other media (and/or storage media) capable of storing, containing, and/or carrying code and/or instructions and/or data.

Compared with the prior art, the invention provides a distributed process monitoring method and device for inter-block collaborative modeling in blocks, which have the following beneficial effects:

1) Based on distributed CCA process monitoring based on partial sub-block communication, combining an SFA method, establishing an SFA local model in each sub-block, establishing a CCA model among the sub-blocks of information interaction, and detecting faults from different characteristic angles of data so as to improve the detection performance and the fault detection rate of the process;

2) The high-dimensional industrial system is decomposed into a plurality of low-dimensional sub-blocks, and a data driving model is respectively built, so that the computational complexity is reduced, the prior knowledge and the process mechanism based division are more reliable, and the state of a large-scale process can be better monitored;

3) The use of sub-block internal modeling is more beneficial to detecting some small faults hidden in high-dimensional data, and the learning capacity is stronger;

4) The fault locating device can provide a preliminary range for fault locating for workers while detecting faults.

While, for purposes of simplicity of explanation, the methodologies are shown and described as a series of acts, it is to be understood and appreciated that the methodologies are not limited by the order of acts, as some acts may, in accordance with one or more embodiments, occur in different orders and/or concurrently with other acts from that shown and described herein or not shown and described herein, as would be understood and appreciated by those skilled in the art.

Those of skill in the art would understand that information, signals, and data may be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips that may be referenced throughout the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof.

Those of skill would further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

The various illustrative logical modules, and circuits described in connection with the embodiments disclosed herein may be implemented or performed with a general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, e.g., a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration.

The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor such the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium may be integral to the processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In the alternative, the processor and the storage medium may reside as discrete components in a user terminal.

As used in this application and in the claims, the terms "a," "an," "the," and/or "the" are not specific to the singular, but may include the plural, unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that the steps and elements are explicitly identified, and they do not constitute an exclusive list, as other steps or elements may be included in a method or apparatus.

The embodiments described above are intended to provide those skilled in the art with a full range of modifications and variations to the embodiments described above without departing from the inventive concept thereof, and therefore the scope of the invention is not limited by the embodiments described above, but is to be accorded the broadest scope consistent with the innovative features recited in the claims.

Claims

1. The distributed process monitoring method for the inter-block collaborative modeling in the block is characterized by comprising two stages of offline modeling and online monitoring, wherein the offline modeling stage specifically comprises the following steps of:

The online monitoring stage specifically comprises the following steps:

2. The method for distributed process monitoring for collaborative modeling between blocks according to claim 1, wherein step S1 further comprises:

X＝[X ₁ ,X ₂ ,…,X _i …,,X _B ]

wherein ,X_i The ith sub-block, and B is the number of sub-blocks.

3. The method for monitoring a distributed process according to claim 1, wherein in step S2, the slow feature extracted by each sub-block corresponds to the expression:

s _i ＝W _i X _i

4. The method for distributed process monitoring in co-modeling among blocks according to claim 1, wherein the statistics in step S3 correspond to the following expressions:

wherein ,

statistics corresponding to the ith sub-block for monitoring the change condition of components representing the essence of the process state in the subspace, s _i Is a slow feature of the ith sub-block.

5. The intra-block inter-block collaborative modeling distributed process monitoring method according to claim 1, wherein the kernel density estimation

The corresponding expression is:

/>

6. The method for monitoring a distributed process according to claim 1, wherein in step S4, a representative correlation analysis is performed between the sub-blocks and the sub-blocks having information interaction with the sub-blocks according to a topology matrix C, and the obtained representative correlation component has the following expression:

7. The method for distributed process monitoring according to claim 1, wherein in step S5, a residual vector is created based on the extracted typical correlation component, and the corresponding expression is:

wherein ,

8. The method for distributed process monitoring of intra-block co-modeling according to claim 1, wherein the final monitoring statistic ET in step S9 ² The corresponding expression is:

wherein ,

final monitoring statistic for ith sub-block, x _new For online data +.>

Conditional probability for an abnormal operating state of the process, +.>

9. The method according to claim 8, wherein in step S9, the control limit CL corresponding to the final monitoring statistic of the ith sub-block _i The prior probability (1-alpha) of abnormal working conditions of the process is given.

10. The method for monitoring a distributed process according to claim 9, wherein in the step S10, the fault detection index FD of the whole process is expressed as follows:

wherein the control limit of the failure detection index FD is 1.

11. A distributed process monitoring device for intra-block inter-block collaborative modeling, comprising:

a memory for storing instructions executable by the processor;

a processor for executing the instructions to implement the method of any one of claims 1-10.

12. A computer readable medium having stored thereon computer instructions, wherein the computer instructions, when executed by a processor, perform the method of any of claims 1-10.