CN116203907B - Chemical process fault diagnosis alarm method and system - Google Patents
Chemical process fault diagnosis alarm method and system Download PDFInfo
- Publication number
- CN116203907B CN116203907B CN202310304189.2A CN202310304189A CN116203907B CN 116203907 B CN116203907 B CN 116203907B CN 202310304189 A CN202310304189 A CN 202310304189A CN 116203907 B CN116203907 B CN 116203907B
- Authority
- CN
- China
- Prior art keywords
- data
- chimpanzee
- fault
- chemical process
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 79
- 238000000034 method Methods 0.000 title claims abstract description 67
- 238000001311 chemical methods and process Methods 0.000 title claims abstract description 51
- 238000012549 training Methods 0.000 claims abstract description 38
- 238000007781 pre-processing Methods 0.000 claims abstract description 9
- 238000012545 processing Methods 0.000 claims abstract description 7
- 241000282577 Pan troglodytes Species 0.000 claims description 66
- 230000008569 process Effects 0.000 claims description 44
- 239000013598 vector Substances 0.000 claims description 36
- 241000282579 Pan Species 0.000 claims description 23
- 238000000354 decomposition reaction Methods 0.000 claims description 21
- 238000005457 optimization Methods 0.000 claims description 19
- 238000012360 testing method Methods 0.000 claims description 13
- 239000011159 matrix material Substances 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 8
- 230000004888 barrier function Effects 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000000739 chaotic effect Effects 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 6
- 230000009471 action Effects 0.000 claims description 3
- 230000006399 behavior Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 abstract description 7
- 238000012824 chemical production Methods 0.000 description 9
- 238000013135 deep learning Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 2
- 239000010723 turbine oil Substances 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Programme-control systems
- G05B19/02—Programme-control systems electric
- G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS], computer integrated manufacturing [CIM]
- G05B19/41875—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS], computer integrated manufacturing [CIM] characterised by quality surveillance of production
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/32—Operator till task planning
- G05B2219/32252—Scheduling production, machining, job shop
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Abstract
The invention discloses a fault diagnosis and alarm method and system for a chemical process. The system consists of a data acquisition module, a data preprocessing module, a model training module, a fault diagnosis module and an alarm module. The data acquisition module acquires the original chemical process data. The data preprocessing module is used for decomposing and dimension-reducing the collected original data set, so that the data utilization rate is improved. The model training module builds a chemical process fault diagnosis model based on STGCN, and trains the built fault diagnosis model through historical data of the chemical process and an improved ChOA algorithm. The fault diagnosis module is used for training a fault model, diagnosing whether faults occur or not, and judging the fault type when the faults are diagnosed. And the alarm module is used for sending out an alarm and displaying the fault type when the fault is diagnosed, so that a factory and workers are reminded of timely processing, the running efficiency of the factory is improved, and the factory maintains a safe and stable production process.
Description
Technical Field
The invention belongs to the field of fault diagnosis of chemical processes, and particularly relates to a fault diagnosis alarm method and system for the chemical processes.
Background
Along with the increasing expansion of the modern industrial production scale and the increasing complexity of process equipment, the safety and stable operation of the chemical production process are also faced with the increasing complexity test. Although the safety management of the chemical production process is continuously improved in recent years in China, and the accident occurrence of the chemical production process is striven for, the chemical production accident still happens. This is mainly affected by two factors: 1. the probability of being smaller under the huge production volume of the second major economic body and the first major industrial country in the world is continuously amplified, so that the chemical production accidents show larger total quantity. 2. The chemical production process is a very complex system, and has the characteristics of large scale, multiple interference, complex fault types, strong real-time performance and the like, and the traditional diagnosis method is difficult to simultaneously meet the accuracy, stability and diagnosis efficiency of diagnosis. And most serious accidents are gradually evolved from tiny problems in production and management, and the teaching and training is not used for reminding all people of the safety problem in the field of chemical production at any time.
The problem of fault diagnosis of industrial processes has been widely studied in the last decades and has achieved great research results. Therefore, a learner classifies these fault diagnosis methods according to different attributes: based on analytical models, qualitative experience, and data driving. In the current big data age background, data-driven fault diagnosis systems are increasingly favored by expert students. In the face of a complex chemical production process, deep learning can process a large amount of nonlinear and multidimensional data more efficiently, and has higher classifying and learning capabilities on the data. Thus, deep learning is becoming a major research direction in data driving.
The existing chemical production process has high data dimension and large data volume, and the diagnosis efficiency of the simple deep learning model is difficult to fully develop. Meanwhile, the deep learning network parameters are complex, and the parameters need to be adjusted according to different data. Therefore, the search for more valuable data and more efficient methods of optimizing parameters is one of the key steps to improve diagnostic efficiency and improve diagnostic accuracy.
Disclosure of Invention
The invention aims to: aiming at the problems pointed out in the background art, the invention discloses a chemical process fault diagnosis alarm method and a system, which build a chemical process fault diagnosis model by utilizing improved ChOA (gas turbine oil turbine) optimized STGCN (gas turbine oil turbine) and aim at improving the utilization of fault information and effectively improving the fault diagnosis efficiency of the chemical process.
The technical scheme is as follows: the invention provides a fault diagnosis and alarm method for a chemical process, which comprises the following steps:
step 1: acquiring an operation variable and a process variable in the Tennex Issman TE process as historical data; according to different fault types, adding fault type labels, and constructing original data sets of different faults;
step 2: decomposing and dimension-reducing the collected original data set; the acquired data sets of different fault types are used as the input of a Multivariate Empirical Mode Decomposition (MEMD), the MEMD decomposition method projects the multivariate input data to a higher dimension, and the signals are decomposed into different components after the multivariate input is cooperatively considered; the decomposed components are further analyzed by adopting kernel principal components to analyze KPCA, important components in the component signals are extracted, and the data dimension is reduced;
step 3: establishing a chemical process fault diagnosis model based on STGCN, and optimizing key parameters of the chemical process fault diagnosis model based on STGCN by utilizing an improved chimpanzee optimization algorithm ChOA; the improved chimpanzee optimization algorithm ChOA uses chaos to initialize population and adds double self-adaptive weight;
step 4: training the established chemical process fault diagnosis model based on the STGCN by utilizing the acquired Tenn Islaman original data set and an improved ChOA algorithm, solving the optimal parameters of the STGCN network and minimizing the loss function error; diagnosing the Tennesie Issmann process data by using the trained and optimized chemical process fault diagnosis model to obtain a diagnosis result and calculate the accuracy rate;
step 5: and (3) judging whether to send out an alarm according to the diagnosis result in the step (4), thereby reminding factories and workers to process in time and displaying the fault type when giving an alarm.
Further, the decomposing step of decomposing the history data into a plurality of different frequencies by using the MEMD decomposition model in the step 2 is as follows:
step 2.1: for the original polynomial process data s (t) = [ s ] 1 (t)，s 2 (t)…，s n (t)] T The projection vector is wherein />Is the kth projection vector along the angle on the unit sphere of (n-1), k=1, 2., K, K is the total number of projection vectors;
step 2.2: after obtaining the projection vector U of the direction vector set, s (t) is calculated along the projection vectorMapping value of (2), recorded as->The calculation formula is as follows:
step 2.3: extracting the mapping signalInstantaneous time when taking local extremum +.>
Step 2.4: for a pair ofInterpolation operation is carried out on the extreme points by using a multi-element spline interpolation method to obtain a multi-element projection envelope line +.> Is->Vector direction signal s l The envelope of (t) is such that, l=1, 2,. -%, n;
step 2.5: calculating the local mean value of the multi-element signals:
step 2.6: calculating the difference d (t) between the multi-element input sequence and the local mean value:
d(t)＝s(t)-m(t)
step 2.7: if d (t) meets the requirement of multi-variable IMF, the condition is marked as d i (t), d i (t) is the component after the ith decomposition, and the component in the original signal is removed to obtain a new original signal k i (t)，i＝1，2，...，m：
k i (t)＝s(t)-d(t)
Step 2.8: repeating the steps 2.2-2.7 until d (t) does not meet the IMF requirement, and recording d (t) as residual error r (t) to obtain final multi-component decomposition signal S (t) = [ k ] 1 (t)，k 2 (t)，...，k m (t)，r(t)]。
Further, the improved chimpanzee optimization algorithm in step 3 includes the steps of:
step 3.1: setting an objective function of the ChOA algorithm as a diagnosis accuracy and initializing related parameters, including: initial position, population scale, iteration number;
step 3.2: in the original chimpanzee algorithm, the initialization is randomly generated according to the dimension and the number of input parameters, so that the logical chaos initialization is introduced in the chimpanzee population initialization process, 4 chimpanzee populations are subjected to wider preliminary search, the search efficiency of the algorithm is improved, and the expression is as follows:
where k (n+1) is the updated individual position and λ is the control variable;
step 3.3: according to the labor division of chimpanzees, the population is divided into 4 classes: common members responsible for driving (Driver) and intercepting (barrer) prey; the primary chase (Chaser) process responsible for young adult chimpanzees; a leader (Attacker) of the prey; chimpanzees have the ability to independently think in a population, and in some cases, confusing hunting behavior;
step 3.4: during the process of chimpanzee hunting, the chimpanzee needs to determine the direction and distance of the next action based on the distance between itself and the prey:
d＝|cx prey (t)-mx chimp (t)|
x chimp (t+1)＝x prey (t)-ad
wherein d is the distance between the prey and the chimpanzee; t is the current number of iterations is the distance between the prey and the chimpanzee; x is x prey (t) is the current location of the prey; x is x chimp (t) is the current location of the chimpanzee; a, m, c are coefficient vectors;
step 3.5: each chimpanzee independently determines the hunting process, i.e. the position vector between each chimpanzee and the prey, according to its own labor division; after the 4 chimpanzees and the prey have determined their position vectors, each chimpanzee updates its position based on the best chimpanzee position and estimates the position of the prey based on the best chimpanzee individual position as follows:
wherein ,dAttacker ,d Barrier ,d Chaser ,d Driver Distance from the prey at the stages of attacking, intercepting, chase and driving the chimpanzee, respectively; x is x Attacker ,x Barrier ,x Chaser ,x Driver Is the vector of the position of the attacking chimpanzee, intercepting chimpanzee, chasing chimpanzee, driving chimpanzee relative to the prey, a 1 ～a 4 ，c 1 ～c 4 ，m 1 ～m 4 Vector coefficients of four chimpanzees, respectively;
step 3.6: in the optimization process, the normal position update of chimpanzees or the position update through a chaotic model is selected, the probability of selection is 50%, and the formula is as follows:
wherein μ is a random number within the range of [0,1 ];
step 3.7: judging whether the maximum iteration times are reached, if not, entering a step 3.2 by the ChOA algorithm; otherwise, ending the operation and outputting a final result.
Further, key parameters of the chemical process fault diagnosis model of the STGCN include: the learning rate, the number of hidden layer nodes and the training iteration number of the STGCN model are used for optimizing key parameters of the chemical process fault diagnosis model based on the STGCN by utilizing an improved chimpanzee optimization algorithm ChOA, and the population is the key parameters to be optimized;
step 4.1: dividing the data in the step 2 into a training set and a testing set for model training and iteratively optimizing key parameters of the STGCN model;
step 4.2: key parameters of the STGCN model are fed into ChOA: the dimension is 3, the dimension is determined according to the kind of key parameters, and the initial value is obtained by chaotic initialization in the step 3.2;
step 4.3: the STGCN model carries out model training according to the training set and the key parameters in the step 4.2, records the diagnosis training result and the accuracy and transmits the diagnosis training result and the accuracy back to the ChOA algorithm, and the accuracy calculation formula is as follows;
the Accuracy Accuracy is the ratio of the fault type of correct classification to the total fault sample ALL, the correct classification refers to the fact that the positive sample TP is searched and the positive sample TN is not searched, the Accuracy is sent to the ChOA and recorded as an adaptability value, and the Accuracy is used as an algorithm iterative optimization index;
step 4.4: repeating the step 4.3, comparing and finding out the optimal fitness value, and recording the key parameters optimized by the iteration until the iteration of the ChOA algorithm is finished;
step 4.5: and (3) taking the optimal key parameters in the step (4.4) as the final use parameters of the model, and sending the final use parameters into a test set for testing to obtain the diagnosis result of the chemical process fault diagnosis model of the final STGCN.
Further, the training of the established STGCN-based chemical process fault diagnosis model by using the collected tennessee iman original data set and the improved ChOA algorithm in the step 4 specifically includes the following steps:
step 4.1: constructing an adjacent matrix A according to the data structure in the step 2, normalizing node parameters of the adjacent matrix A, and forming graph data needed by a model:
wherein ,representing node x i Mean, sigma of i Representing the standard deviation of the node.
Step 4.2: constructing a multi-node graph G according to chemical process data and time relation i ：
G i ＝(x i ,E,A)
Wherein E represents a set of edges between nodes;
step 4.3: a time convolution module is formed by a gating linear unit GLU and a one-dimensional convolution network and is used for capturing time characteristics; the gating linear unit can select to transfer the needed information to the next node, and the expression formula is as follows:
where M and N are different convolution kernels, c 1 and c2 Is a different bias parameter;
step 4.4: the graph convolution module is used for extracting high-order signs on a space domain, the relevance and the global property of fault data in a chemical process are fully utilized, and a Chebyshev polynomial approximation is adopted in convolution, and a convolution formula is as follows:
where Z is the graph convolution kernel size, T Z Is a polynomial expansion approximation of the Laplace matrix, Θ z Is a polynomial coefficient and the final graph convolution can be expressed as:
wherein ,Di and D0 The size of the feature map is input and output, and D represents dimension features;
step 4.5: and finally, performing inverse normalization processing on the obtained data, and outputting the fault type according to the test result.
The invention also discloses a chemical process fault diagnosis alarm system, which comprises a data acquisition module, a data preprocessing module, a model training module, a fault diagnosis module and an alarm module;
the data acquisition module is used for acquiring an operation variable and a process variable in the Tennex Issmann TE process as historical data; according to different fault types, adding fault type labels, and constructing original data sets of different faults;
the data preprocessing module is used for decomposing and dimension-reducing the acquired original data set; the acquired data sets of different fault types are used as the input of a Multivariate Empirical Mode Decomposition (MEMD), the MEMD decomposition method projects the multivariate input data to a higher dimension, and the signals are decomposed into different components after the multivariate input is cooperatively considered; the decomposed components are further analyzed by adopting kernel principal components to analyze KPCA, important components in the component signals are extracted, and the data dimension is reduced;
the model training module is used for establishing a chemical process fault diagnosis model based on the STGCN, training the established chemical process fault diagnosis model based on the STGCN through the acquired Tennex Issman original data set and an improved ChOA algorithm, solving the optimal parameters of the network and minimizing the loss function error;
and the fault diagnosis module is used for diagnosing the Tenn Islaman process data by using the trained and optimized chemical process fault diagnosis model to obtain a diagnosis result and judging the fault type when the fault is diagnosed.
And the alarm module is used for sending out an alarm and displaying the fault type when the fault is diagnosed, so that a factory and workers are reminded of timely processing.
The beneficial effects are that:
1. the invention uses MEMD decomposition and KPCA to reconstruct the input signal, removes the influence of noise on model precision and effectively reduces the input dimension. 2. Aiming at the defect that the ChOA algorithm is easy to fall into local optimum, the invention provides an improved ChOA algorithm, and the optimization capacity of the ChOA algorithm is enhanced and the diagnosis performance of a model is improved by using chaos to initialize a population and adding double self-adaptive weights. 3. The invention utilizes the improved ChOA algorithm to synchronously optimize the MEMD and the STGCN, and captures the potential relation between the input characteristic factors and the model parameters better.
Drawings
FIG. 1 is a system flow diagram of the present invention;
FIG. 2 is a flow chart of the proposed fault diagnosis method;
fig. 3 is a structural diagram of STGCN;
fig. 4 is a schematic diagram of TE process.
Detailed Description
Embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.
The invention provides a chemical process fault diagnosis alarm system, which is provided with a chemical process fault diagnosis alarm method, as shown in figure 1, and specifically comprises a data acquisition module, a data preprocessing module, a model training module, a fault diagnosis module and an alarm module.
The data acquisition module is used for acquiring operation variables and process variables in the Tennex Issmann (TE) production process as historical data; and adding fault type labels according to different fault types, and constructing original data sets of different faults. Referring to fig. 4, the Tennessee Eastern (TE) production process is a conventional production process in the chemical field, and specific processes are not described herein, wherein a dotted line is a control loop, and a solid line is a pipeline.
There are 21 different types of faults in the TE process: the faults 1 to 15 and 21 are preset fault types, and the faults 16 to 20 are unknown faults caused by uncontrollable factors.
The data preprocessing module is used for decomposing and dimension-reducing the acquired original data set; the acquired data sets of different fault types are used as the input of multi-element empirical mode decomposition (MEMD), the MEMD decomposition method can project multi-element input data to a higher dimension, and the signals are decomposed into different components after the multi-element input is considered cooperatively; the decomposed components are further subjected to Kernel Principal Component Analysis (KPCA), important components in the component signals are extracted, and the data dimension is reduced.
The decomposition steps of the MEMD decomposition model are as follows:
for the original polynomial process data s (t) = [ s ] 1 (t)，s 2 (t)…，s n (t)] T The projection vector is wherein />Is the kth projection vector along the angle on the unit sphere of (n-1), k=1, 2.
After obtaining the projection vector U of the direction vector set, s (t) is calculated along the projection vectorIs recorded as the mapping value of (2)The calculation formula is as follows:
extracting the mapping signalInstantaneous time when taking local extremum +.>
For a pair ofInterpolation operation is carried out on the extreme points by using a multi-element spline interpolation method to obtain a multi-element projection envelope line +.> Is->Vector direction signal s l Envelope of (t), l=1, 2,..n.
Calculating the local mean value of the multi-element signals:
calculating the difference d (t) between the multi-element input sequence and the local mean value:
d(t)＝s(t)-m(t)
if d (t) meets the requirement of multi-variable IMF, the condition is marked as d i (t), d i (t) is the component after the ith decomposition. And removing the component in the original signal to obtain a new original signal k i (t)，i＝1，2，...，m：
k i (t)＝s(t)-d(t)
Repeating the steps until d (t) does not meet the requirement of IMF, and recording d (t) as residual error r (t). Obtaining a final multivariate decomposition signal S (t) = [ k ] 1 (t)，k 2 (t)，...，k m (t)，r(t)]。
The model training module is used for establishing a chemical process fault diagnosis model based on the STGCN, training the established chemical process fault diagnosis model based on the STGCN through the acquired Tenn Islaman original data set and an improved chimpanzee optimization algorithm ChOA, solving the optimal key parameters of the network and minimizing the loss function error.
The fault diagnosis module is used for diagnosing the tennessee Issmann process data by using the trained and optimized chemical process fault diagnosis model to obtain a diagnosis result, and judging the fault type when the fault is diagnosed. And finally, performing inverse normalization processing on the obtained data, and outputting the fault type according to the test result.
The STGCN model comprises the following specific steps:
and 2, constructing an adjacent matrix A according to the data structure in the step 2, and normalizing node parameters of the adjacent matrix A to form graph data required by the model.
wherein ,representing node x i Mean, sigma of i Representing the standard deviation of the node.
Constructing a multi-node graph G according to chemical process data and time relation i 。
G i ＝(x i ,E,A)
Where E represents the set of edges between nodes.
A time convolution module is formed by a Gated Linear Unit (GLU) and a one-dimensional convolution network for capturing time characteristics. The gating linear unit can select to transfer the needed information to the next node, and the expression formula is as follows:
where M and N are different convolution kernels, c 1 and c2 Are different bias parameters.
The graph convolution module is used for extracting high-order signs on a spatial domain, and the relevance and the global property of the fault data of the chemical process are fully utilized. In the convolution, chebyshev polynomial approximation is adopted, and the convolution formula is as follows:
where Z is the graph convolution kernel size, T Z Is a polynomial expansion approximation of the Laplace matrix, Θ z Is a polynomial coefficient and the final graph convolution can be expressed as:
wherein ,Di and D0 Is the size of the input and output feature map, D represents the dimension feature.
The improved ChOA algorithm comprises the following steps:
setting an objective function of the ChOA algorithm as a diagnosis accuracy and initializing related parameters, including: initial position, population size, iteration number. The parameters to be optimized are input into the improved chimpanzee optimization algorithm and then exist in the form of population, so that the population scale and the iteration number of the optimized real content are selected by experience.
In the original chimpanzee algorithm, the initialization is randomly generated from the dimensions and number of input parameters. Therefore, the Logistic chaos initialization is introduced in the chimpanzee population initialization process, 4 chimpanzee populations can be subjected to wider preliminary search, the search efficiency of an algorithm is improved, and the expression is as follows:
where k (n+1) is the updated individual position and λ is the control variable.
According to the labor division of chimpanzees, the population is divided into 4 classes: common members responsible for driving (Driver) and intercepting (barrer) prey; the primary chase (Chaser) process responsible for young adult chimpanzees; a leader (Attacker) of hunting. Chimpanzees have the ability to independently think in a population, and in some cases, confusing hunting behavior.
During the process of chimpanzee hunting, the chimpanzee needs to determine the direction and distance of the next action based on the distance between itself and the prey.
d＝|cx prey (t)-mx chimp (t)|
x chimp (t+1)＝x prey (t)-ad
Wherein d is the distance between the prey and the chimpanzee; t is the current number of iterations is the distance between the prey and the chimpanzee; x is x prey (t) is the current location of the prey; x is x chimp (t) is a chimpanzeeA current location; a, m, c are coefficient vectors. Since the aggregation of chimpanzees is different for each class, its coefficient vector update needs to be iterated using different formulas.
Each chimpanzee independently determines the course of the hunting, i.e. the position vector between each chimpanzee and the prey, based on its own labor division. After the 4 chimpanzees and the prey have determined their position vectors, each chimpanzee updates its position based on the best chimpanzee position and estimates the position of the prey based on the best chimpanzee individual position as follows:
wherein ,dAttacker ,d Barrier ,d Chaser ,d Driver Distance from the prey at the stages of attacking, intercepting, chase and driving the chimpanzee, respectively; x is x Attacker ,x Barrier ,x Chaser ,x Driver Is a vector of the positions of the chimpanzee relative to the prey that attacks the chimpanzee, intercepts the chimpanzee, chases the chimpanzee, and drives the chimpanzee. a, a 1 ～a 4 ，c 1 ～c 4 ，m 1 ～m 4 The vector coefficients of the four chimpanzees are respectively.
Driven by instinct and other factors, chimpanzees may also break through the current territory to hunting, avoiding ChOA from falling into local optima and slow convergence when solving high-dimensional problems. In the optimization process, the normal position update of chimpanzees or the position update through a chaotic model is selected, the probability of selection is 50%, and the formula is as follows:
wherein μ is a random number in the range of [0,1 ].
Judging whether the maximum iteration times are reached, if not, entering a chimpanzee population initialization process by the ChOA algorithm; otherwise, ending the operation and outputting a final result.
The learning rate, the number of hidden layer nodes and the training iteration number of the STGCN model are used for optimizing key parameters of the chemical process fault diagnosis model based on the STGCN by utilizing an improved chimpanzee optimization algorithm ChOA, and the population is the key parameters to be optimized, and specifically comprises the following steps:
step 4.1: dividing the data after the data preprocessing module into a training set and a testing set for model training and key parameters of the iterative optimization STGCN model.
Step 4.2: key parameters of the STGCN model are fed into ChOA: the dimension is 3, and the dimension is determined according to the type of the key parameters, in this embodiment, the key parameters are the learning rate, the number of hidden layer nodes and the training iteration number, and the initial value is obtained by chaotic initialization.
Step 4.3: the STGCN model carries out model training according to the training set and the key parameters in the step 4.2, records the diagnosis training result and the accuracy and transmits the diagnosis training result and the accuracy back to the ChOA algorithm, and the accuracy calculation formula is as follows;
the Accuracy Accuracy is the ratio of the fault type of correct classification to the total fault sample ALL, the correct classification refers to the fact that the positive sample TP is searched and the positive sample TN is not searched, and the Accuracy is sent to the ChOA and recorded as an adaptability value and used as an algorithm iterative optimization index.
Step 4.4: repeating the step 4.3, comparing and finding out the optimal fitness value, and recording the key parameters optimized by the iteration until the iteration of the ChOA algorithm is finished;
step 4.5: and (3) taking the optimal key parameters in the step (4.4) as the final use parameters of the model, and sending the final use parameters into a test set for testing to obtain the diagnosis result of the chemical process fault diagnosis model of the final STGCN.
And the alarm module is used for sending out an alarm and displaying the fault type when the fault is diagnosed, so that a factory and workers are reminded of timely processing.
The present invention is not limited to the above-described embodiments, and any equivalent or modified embodiments according to the technical solution of the present invention and the inventive concept thereof are included in the scope of the present invention within the knowledge of those skilled in the art.
Claims (4)
1. The fault diagnosis and alarm method for the chemical process is characterized by comprising the following steps:
step 1: acquiring an operation variable and a process variable in the Tennex Issman TE process as historical data; according to different fault types, adding fault type labels, and constructing original data sets of different faults;
step 2: decomposing and dimension-reducing the collected original data set; the acquired data sets of different fault types are used as the input of a Multivariate Empirical Mode Decomposition (MEMD), the MEMD decomposition method projects the multivariate input data to a higher dimension, and the signals are decomposed into different components after the multivariate input is cooperatively considered; the decomposed components are further analyzed by adopting kernel principal components to analyze KPCA, important components in the component signals are extracted, and the data dimension is reduced;
step 3: establishing a chemical process fault diagnosis model based on STGCN, and optimizing key parameters of the chemical process fault diagnosis model based on STGCN by utilizing an improved chimpanzee optimization algorithm ChOA; training the established chemical process fault diagnosis model based on the STGCN by utilizing the acquired Tenn Islaman original data set and an improved ChOA algorithm, solving the optimal key parameters of the STGCN network and minimizing the loss function error; the improved chimpanzee optimization algorithm ChOA uses chaos to initialize population and adds double self-adaptive weight; the improved chimpanzee optimization algorithm comprises the following steps:
step 3.1: setting an objective function of the ChOA algorithm as a diagnosis accuracy and initializing related parameters, including: initial position, population scale, iteration number;
step 3.2: in the original chimpanzee algorithm, the initialization is randomly generated according to the dimension and the number of input parameters, so that the logical chaos initialization is introduced in the chimpanzee population initialization process, 4 chimpanzee populations are subjected to wider preliminary search, the search efficiency of the algorithm is improved, and the expression is as follows:
where k (n+1) is the updated individual position and λ is the control variable;
step 3.3: according to the labor division of chimpanzees, the population is divided into 4 classes: attack chimpanzees, intercept chimpanzees, chase chimpanzees, and repel chimpanzees; chimpanzees have the ability to independently think in a population, and in some cases, confusing hunting behavior;
step 3.4: during the process of chimpanzee hunting, the chimpanzee needs to determine the direction and distance of the next action based on the distance between itself and the prey:
d＝|cx prey (t)-mx chimp (t)|
x chimp (t+1)＝x prey (t)-ad
wherein d is the distance between the prey and the chimpanzee; t is the current number of iterations is the distance between the prey and the chimpanzee; x is x prey (t) is the current location of the prey; x is x chimp (t) is the current location of the chimpanzee; a, m, c are coefficient vectors;
step 3.5: each chimpanzee independently determines the hunting process, i.e. the position vector between each chimpanzee and the prey, according to its own labor division; after the 4 chimpanzees and the prey have determined their position vectors, each chimpanzee updates its position based on the best chimpanzee position and estimates the position of the prey based on the best chimpanzee individual position as follows:
wherein ,dAttacker ,d Barrier ,d Chaser ,d Driver Representing the distance from the prey during the stages of attack, interception, chase and driving of the chimpanzee, respectively; x is x Attacker ,x Barrier ,x Chaser ,x Driver Is the vector of the position of the attacking chimpanzee, intercepting chimpanzee, chasing chimpanzee, driving chimpanzee relative to the prey, a 1 ～a 4 ，c 1 ～c 4 ，m 1 ～m 4 Vector coefficients of four chimpanzees, respectively;
step 3.6: in the optimization process, the normal position update of chimpanzees or the position update through a chaotic model is selected, the probability of selection is 50%, and the formula is as follows:
wherein μ is a random number within the range of [0,1 ];
step 3.7: judging whether the maximum iteration times are reached, if not, entering a step 3.2 by the ChOA algorithm; otherwise, ending the operation and outputting a final result;
step 4: diagnosing the Tennesie Issmann process data by using the trained and optimized chemical process fault diagnosis model to obtain a diagnosis result and calculate the accuracy rate;
step 4.1: constructing an adjacent matrix A according to the data structure in the step 2, normalizing node parameters of the adjacent matrix A, and forming graph data needed by a model:
wherein ,representing node x i Mean, sigma of i Representing the standard deviation of the node;
step 4.2: constructing a multi-node graph G according to chemical process data and time relation i ：
G i ＝(x i ,E,A)
Wherein E represents a set of edges between nodes;
step 4.3: a time convolution module is formed by a gating linear unit GLU and a one-dimensional convolution network and is used for capturing time characteristics; the gating linear unit can select to transfer the needed information to the next node, and the expression formula is as follows:
where M and N are different convolution kernels, c 1 and c2 Is a different bias parameter;
step 4.4: the graph convolution module is used for extracting high-order signs on a space domain, the relevance and the global property of fault data in a chemical process are fully utilized, and a Chebyshev polynomial approximation is adopted in convolution, and a convolution formula is as follows:
where Z is the graph convolution kernel size, T Z Is a polynomial expansion approximation of the Laplace matrix, Θ z Is a polynomial coefficientThe final graph convolution can be expressed as:
wherein ,Di and D0 The size of the input and output feature graphs is that D represents dimension features and R is a real number set;
step 4.5: finally, performing inverse normalization processing on the obtained data, and outputting a fault type according to a test result;
step 5: and (3) judging whether to send out an alarm according to the diagnosis result in the step (4), thereby reminding factories and workers to process in time and displaying the fault type when giving an alarm.
2. The method for diagnosing and alarming a fault in a chemical process according to claim 1, wherein the decomposing step of decomposing the history data into a plurality of different frequencies by using the MEMD decomposition model in the step 2 is as follows:
step 2.1: for the original polynomial process data s (t) = [ s ] 1 (t)，s 2 (t)…，s n (t)] T The projection vector is wherein />Is the kth projection vector along the angle on the unit sphere of (n-1), k=1, 2., K, K is the total number of projection vectors;
step 2.2: after obtaining the projection vector U of the direction vector set, s (t) is calculated along the projection vectorMapping value of (2), recorded as->The calculation formula is as follows:
step 2.3: extracting the mapping signalInstantaneous time when taking local extremum +.>
Step 2.4: for a pair ofInterpolation operation is carried out on the extreme points by using a multi-element spline interpolation method to obtain a multi-element projection envelope line +.> Is->Vector direction signal s l The envelope of (t) is such that, l=1, 2,. -%, n;
step 2.5: calculating the local mean value of the multi-element signals:
step 2.6: calculating the difference d (t) between the multi-element input sequence and the local mean value:
d(t)＝s(t)-m(t)
step 2.7: if d (t) meets the requirement of multi-variable IMF, the condition is marked as d i (t), d i (t) is the component after the ith decomposition and the component in the original signalThe quantitative components are removed to obtain a new original signal k i (t)，i＝1，2，...，m：
k i (t)＝s(t)-d(t)
Step 2.8: repeating the steps 2.2-2.7 until d (t) does not meet the IMF requirement, and recording d (t) as residual error r (t) to obtain final multi-component decomposition signal S (t) = [ k ] 1 (t)，k 2 (t)，...，k m (t)，r(t)]。
3. The chemical process fault diagnosis alarm method according to claim 1, wherein the key parameters of the chemical process fault diagnosis model of the STGCN include: the learning rate, the number of hidden layer nodes and the training iteration number of the STGCN model are used for optimizing key parameters of the chemical process fault diagnosis model based on the STGCN by utilizing an improved chimpanzee optimization algorithm ChOA, and the population is the key parameters to be optimized;
step 4.1: dividing the data in the step 2 into a training set and a testing set for model training and iteratively optimizing key parameters of the STGCN model;
step 4.2: key parameters of the STGCN model are fed into ChOA: the dimension is 3, the dimension is determined according to the kind of key parameters, and the initial value is obtained by chaotic initialization in the step 3.2;
step 4.3: the STGCN model carries out model training according to the training set and the key parameters in the step 4.2, records the diagnosis training result and the accuracy and transmits the diagnosis training result and the accuracy back to the ChOA algorithm, and the accuracy calculation formula is as follows;
the Accuracy Accuracy is the ratio of the fault type of correct classification to the total fault sample ALL, the correct classification refers to the fact that the positive sample TP is searched and the positive sample TN is not searched, the Accuracy is sent to the ChOA and recorded as an adaptability value, and the Accuracy is used as an algorithm iterative optimization index;
step 4.4: repeating the step 4.3, comparing and finding out the optimal fitness value, and recording the key parameters optimized by the iteration until the iteration of the ChOA algorithm is finished;
step 4.5: and (3) taking the optimal key parameters in the step (4.4) as the final use parameters of the model, and sending the final use parameters into a test set for testing to obtain the diagnosis result of the chemical process fault diagnosis model of the final STGCN.
4. A system based on the chemical process fault diagnosis and alarm method as claimed in any one of claims 1 to 3, characterized in that the system comprises a data acquisition module, a data preprocessing module, a model training module, a fault diagnosis module and an alarm module;
the data acquisition module is used for acquiring an operation variable and a process variable in the Tennex Issmann TE process as historical data; according to different fault types, adding fault type labels, and constructing original data sets of different faults;
the data preprocessing module is used for decomposing and dimension-reducing the acquired original data set; the acquired data sets of different fault types are used as the input of a Multivariate Empirical Mode Decomposition (MEMD), the MEMD decomposition method projects the multivariate input data to a higher dimension, and the signals are decomposed into different components after the multivariate input is cooperatively considered; the decomposed components are further analyzed by adopting kernel principal components to analyze KPCA, important components in the component signals are extracted, and the data dimension is reduced;
the model training module is used for modeling a chemical process fault diagnosis model based on the STGCN, training the established chemical process fault diagnosis model based on the STGCN through the acquired Tennex Issman original data set and an improved ChOA algorithm, solving the optimal key parameters of the network and minimizing the loss function error;
the fault diagnosis module is used for diagnosing the Tenn Islaman process data by using the trained and optimized chemical process fault diagnosis model to obtain a diagnosis result and judging the fault type when a fault is diagnosed;
and the alarm module is used for sending out an alarm and displaying the fault type when the fault is diagnosed, so that a factory and workers are reminded of timely processing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310304189.2A CN116203907B (en) | 2023-03-27 | 2023-03-27 | Chemical process fault diagnosis alarm method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310304189.2A CN116203907B (en) | 2023-03-27 | 2023-03-27 | Chemical process fault diagnosis alarm method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116203907A CN116203907A (en) | 2023-06-02 |
| CN116203907B true CN116203907B (en) | 2023-10-20 |
Family
ID=86517370
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310304189.2A Active CN116203907B (en) | 2023-03-27 | 2023-03-27 | Chemical process fault diagnosis alarm method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116203907B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6728746B1 (en) * | 1995-02-14 | 2004-04-27 | Fujitsu Limited | Computer system comprising a plurality of machines connected to a shared memory, and control method for a computer system comprising a plurality of machines connected to a shared memory |
| CN108536130A (en) * | 2018-05-14 | 2018-09-14 | 浙江大学 | A kind of Fault Diagnosis in Chemical Process system of colony intelligence optimizing |
| CN108536107A (en) * | 2018-05-14 | 2018-09-14 | 浙江大学 | Colony intelligence optimizing fault diagnosis system based on hybrid optimized parameter |
| CN108764305A (en) * | 2018-05-14 | 2018-11-06 | 浙江大学 | A kind of improved colony intelligence machine learning fault diagnosis system |
| CN113614831A (en) * | 2019-03-22 | 2021-11-05 | 英芙勒玛提克斯公司 | System and method for deriving and optimizing classifiers from multiple data sets |
| CN114330671A (en) * | 2022-01-06 | 2022-04-12 | 重庆大学 | Traffic flow prediction method based on Transformer space-time diagram convolution network |
| CN115115389A (en) * | 2022-03-11 | 2022-09-27 | 南京邮电大学 | Express customer loss prediction method based on value subdivision and integrated prediction |
| CN115358291A (en) * | 2022-07-21 | 2022-11-18 | 国网宁夏电力有限公司电力科学研究院 | Transformer fault diagnosis method, medium and system |
| CN115457307A (en) * | 2022-08-10 | 2022-12-09 | 淮阴工学院 | Chemical process fault diagnosis method based on improved residual error network |
| CN115561005A (en) * | 2022-10-25 | 2023-01-03 | 淮阴工学院 | Chemical process fault diagnosis method based on EEMD decomposition and lightweight neural network |
| CN115755219A (en) * | 2022-10-18 | 2023-03-07 | 长江水利委员会水文局 | Flood forecast error real-time correction method and system based on STGCN |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9529974B2 (en) * | 2008-02-25 | 2016-12-27 | Georgetown University | System and method for detecting, collecting, analyzing, and communicating event-related information |
| US10424394B2 (en) * | 2011-10-06 | 2019-09-24 | Sequenom, Inc. | Methods and processes for non-invasive assessment of genetic variations |
-
2023
- 2023-03-27 CN CN202310304189.2A patent/CN116203907B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6728746B1 (en) * | 1995-02-14 | 2004-04-27 | Fujitsu Limited | Computer system comprising a plurality of machines connected to a shared memory, and control method for a computer system comprising a plurality of machines connected to a shared memory |
| CN108536130A (en) * | 2018-05-14 | 2018-09-14 | 浙江大学 | A kind of Fault Diagnosis in Chemical Process system of colony intelligence optimizing |
| CN108536107A (en) * | 2018-05-14 | 2018-09-14 | 浙江大学 | Colony intelligence optimizing fault diagnosis system based on hybrid optimized parameter |
| CN108764305A (en) * | 2018-05-14 | 2018-11-06 | 浙江大学 | A kind of improved colony intelligence machine learning fault diagnosis system |
| CN113614831A (en) * | 2019-03-22 | 2021-11-05 | 英芙勒玛提克斯公司 | System and method for deriving and optimizing classifiers from multiple data sets |
| CN114330671A (en) * | 2022-01-06 | 2022-04-12 | 重庆大学 | Traffic flow prediction method based on Transformer space-time diagram convolution network |
| CN115115389A (en) * | 2022-03-11 | 2022-09-27 | 南京邮电大学 | Express customer loss prediction method based on value subdivision and integrated prediction |
| CN115358291A (en) * | 2022-07-21 | 2022-11-18 | 国网宁夏电力有限公司电力科学研究院 | Transformer fault diagnosis method, medium and system |
| CN115457307A (en) * | 2022-08-10 | 2022-12-09 | 淮阴工学院 | Chemical process fault diagnosis method based on improved residual error network |
| CN115755219A (en) * | 2022-10-18 | 2023-03-07 | 长江水利委员会水文局 | Flood forecast error real-time correction method and system based on STGCN |
| CN115561005A (en) * | 2022-10-25 | 2023-01-03 | 淮阴工学院 | Chemical process fault diagnosis method based on EEMD decomposition and lightweight neural network |
Non-Patent Citations (4)
| Title |
|---|
| 基于VAE-DBN的故障分类方法在化工过程中的应用;张祥;崔哲;董玉玺;田文德;;过程工程学报(03);全文 * |
| 张祥 ; 崔哲 ; 董玉玺 ; 田文德 ; .基于VAE-DBN的故障分类方法在化工过程中的应用.过程工程学报.2018,(03),全文. * |
| 支持向量机在化工过程故障诊断中的应用;蒋强;黄剑;;沈阳理工大学学报(05);全文 * |
| 蒋强 ; 黄剑 ; .支持向量机在化工过程故障诊断中的应用.沈阳理工大学学报.2016,(05),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116203907A (en) | 2023-06-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110361176B (en) | Intelligent fault diagnosis method based on multitask feature sharing neural network | |
| CN109597401B (en) | Equipment fault diagnosis method based on data driving | |
| CN109657945B (en) | Industrial production process fault diagnosis method based on data driving | |
| CN111459144A (en) | Airplane flight control system fault prediction method based on deep cycle neural network | |
| Du et al. | GAN-based anomaly detection for multivariate time series using polluted training set | |
| CN115018021A (en) | Machine room abnormity detection method and device based on graph structure and abnormity attention mechanism | |
| CN111340063A (en) | Coal mill data anomaly detection method | |
| CN105607631B (en) | The weak fault model control limit method for building up of batch process and weak fault monitoring method | |
| Wu et al. | A transformer-based approach for novel fault detection and fault classification/diagnosis in manufacturing: A rotary system application | |
| CN111507365A (en) | Confidence rule automatic generation method based on fuzzy clustering | |
| CN114548591A (en) | Time sequence data prediction method and system based on hybrid deep learning model and Stacking | |
| Zhang et al. | MS-TCN: A multiscale temporal convolutional network for fault diagnosis in industrial processes | |
| Tripathy et al. | Explaining Anomalies in Industrial Multivariate Time-series Data with the help of eXplainable AI | |
| CN114443338A (en) | Sparse negative sample-oriented anomaly detection method, model construction method and device | |
| CN115905848A (en) | Chemical process fault diagnosis method and system based on multi-model fusion | |
| CN116203907B (en) | Chemical process fault diagnosis alarm method and system | |
| CN117333038A (en) | Economic trend analysis system based on big data | |
| Jin et al. | Transfer learning with quantile regression | |
| CN111553115A (en) | Large-span bridge vibration response prediction method under typhoon action based on data driving | |
| CN116226739A (en) | Map convolution network industrial process fault diagnosis method based on space-time fusion | |
| CN114841063A (en) | Aero-engine residual life prediction method based on deep learning | |
| Hallin et al. | Multivariate signed-rank tests in vector autoregressive order identification | |
| Chernyshev et al. | Integration of building information modeling and artificial intelligence systems to create a digital twin of the construction site | |
| Bewley et al. | Learning Interpretable Models of Aircraft Handling Behaviour by Reinforcement Learning from Human Feedback | |
| Li et al. | A new random forest method based on belief decision trees and its application in intention estimation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |