CN114021469B - Method for monitoring one-stage furnace process based on mixed sequence network - Google Patents
Method for monitoring one-stage furnace process based on mixed sequence network Download PDFInfo
- Publication number
- CN114021469B CN114021469B CN202111348802.8A CN202111348802A CN114021469B CN 114021469 B CN114021469 B CN 114021469B CN 202111348802 A CN202111348802 A CN 202111348802A CN 114021469 B CN114021469 B CN 114021469B
- Authority
- CN
- China
- Prior art keywords
- network
- output
- hidden layer
- decoder
- sample
- 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
- 238000000034 method Methods 0.000 title claims abstract description 106
- 230000008569 process Effects 0.000 title claims abstract description 74
- 238000012544 monitoring process Methods 0.000 title claims abstract description 41
- 238000001514 detection method Methods 0.000 claims abstract description 33
- 239000010410 layer Substances 0.000 claims description 49
- 238000013507 mapping Methods 0.000 claims description 20
- 238000012549 training Methods 0.000 claims description 20
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000005065 mining Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 239000002356 single layer Substances 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 4
- 238000003062 neural network model Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 12
- 238000002474 experimental method Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- QGZKDVFQNNGYKY-UHFFFAOYSA-N Ammonia Chemical compound N QGZKDVFQNNGYKY-UHFFFAOYSA-N 0.000 description 2
- 125000004122 cyclic group Chemical group 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000010978 in-process monitoring Methods 0.000 description 2
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 239000002994 raw material Substances 0.000 description 2
- 229910021529 ammonia Inorganic materials 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000001257 hydrogen Substances 0.000 description 1
- 229910052739 hydrogen Inorganic materials 0.000 description 1
- 125000004435 hydrogen atom Chemical class [H]* 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000003345 natural gas Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/10—Numerical modelling
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a method for monitoring a one-stage furnace process based on a mixed sequence network, wherein the network structure comprises an encoder, a modal identification network and a decoder, the encoder consists of a group of RNNs, and the last hidden layer output of the encoder is used for reconstructing a sample at the last moment of a sequence through the decoder. The decoder is composed of a plurality of sub-decoder parts, and the final reconstruction value is obtained by weighting the weight output by the modal identification network. The network parameters are trained by adding the weight reconstruction errors, and the information entropy aiming at the weight is added into the loss function, so that a more accurate mode identification effect is obtained, and the network is prevented from collapsing towards a single mode. Finally, based on the designed neural network model, a weighted square prediction error is constructed to indicate whether a fault occurs in the process, and fault variables are identified through contribution. The method can accurately monitor the process of the primary furnace, and has high fault detection and identification accuracy.
Description
Technical Field
The invention belongs to the field of fault detection and fault identification of a primary furnace process, and particularly relates to a method for monitoring the primary furnace process based on a mixed sequence network.
Background
In order to maintain stable operation of a production process, a process monitoring technology receives great importance over the years, fault detection and fault identification are main components in process monitoring, the purpose of fault detection is to detect whether a fault exists in an industrial process which is currently operated, and fault identification is to locate fault variables after determining that a fault occurs so as to help engineers to conduct fault detection and recovery.
The primary furnace is a core production device in the process of synthesizing ammonia, and mainly converts natural gas into raw material hydrogen, the reaction steps of the process are complex, and a strong nonlinear relation often exists among process variables, so that the traditional data driving fault detection method based on linear assumption cannot show good detection performance in the process. In addition, due to operating condition variations and process raw material variations, a segment of furnace process data often exhibits multi-modal characteristics, which also makes it difficult for conventional data-driven models under single operating conditions or steady-state assumptions to effectively perform process monitoring. In recent years, deep learning has become a research hotspot, which mainly includes some neural network-based feature extraction models, and has been widely applied to various fields related to industry due to its excellent performance in processing nonlinear data.
The self-encoder is a commonly used deep learning model, and can be used for mining complex features in data, but the self-encoder cannot mine dynamic features in the data, and in addition, the self-encoder is directly used for aiming at multi-modal features shown in the data, so that confusion among the multi-modal features can be possibly caused, and the monitoring performance is reduced.
Disclosure of Invention
In order to solve the problems of nonlinearity, multiple modes and the like in the one-stage furnace process, the invention provides a method for monitoring the one-stage furnace process based on a mixed sequence network.
The specific technical scheme of the invention is as follows:
a method for performing a stage of furnace process monitoring based on a mixed sequence network, the method comprising the steps of:
S1: constructing a process monitoring model based on a mixed sequence network, and performing feature mining on a primary furnace process;
the process monitoring model comprises an encoder, a modal identification network and a decoder; the encoder is RNN and is used for dynamic characteristics in the process of mining; the mode identification network outputs a group of weights through a softmax layer after passing through a hidden layer after being connected with the hidden layer output at the last moment of the encoder, and the weights are used for indicating the mode of the current sequence sample; the decoder is also connected with the hidden layer output at the last moment of the encoder and is used for reconstructing the sample at the last moment in the input sequence; the decoder consists of a plurality of single-layer neural networks, and the final reconstruction value is obtained after each neural network output is weighted and summed through the weight output by the modal identification network;
s2: collecting process data under normal working conditions of a section of furnace process, constructing a data set, setting a sequence length L at the same time, serializing the data set, and taking the obtained sequence data set as a training data set X for model training; wherein the nth input sequence is
S3: inputting the training data set X into a process monitoring model based on a mixed sequence network, performing forward propagation to obtain a reconstruction value, and minimizing a loss function by an iterative method until model parameters converge or the maximum iterative times are reached to obtain a trained process monitoring model;
S4: calculating a detection index WSPE by using training data, and calculating a control limit con wspe by using a nuclear density estimation method;
S5: constructing an input sequence of a required length L by utilizing a section of furnace on-line detection sample x and samples at the previous L-1 moments, substituting the input sequence into a trained process monitoring model in S4 to obtain reconstruction output of a decoder of x and output p= [ p 1 p2 … pK ] of a modal identification network; the reconstructed output of the ith sub-decoder is noted as
S6: calculating a detection index WSPE o by using the online sample and the reconstruction output thereof, and comparing the detection index with a control limit con wspe, wherein the online sample is a normal sample when WSPE o≤conwspe; when WSPE o＞conwspe, the current sample is considered as a fault sample, and the fault sample is further subjected to fault identification; let x= [ x 1 x2 … xm ] be the line sample, the reconstruction value of the ith decoder after the model is taken asThe calculation method of the contribution index of the j-th variable is as follows:
s7: and regarding the variable with higher contribution degree as a fault variable according to the requirement.
Further, the step S3 is implemented by the following substeps:
(1) X n is brought into a process monitoring model constructed by S1, and X n is transmitted forward to obtain hidden layer output of RNN
Wherein U e denotes the weights of hidden layer features mapping inputs to the RNN,M is the variable number of the input samples; w e denotes the weight to map the hidden layer output at time t-1 in the RNN to the hidden layer output at time t in the RNN,H e is the node number of the RNN hidden layer; /(I)And/>Hidden layer output representing time t and time t-1 respectively,/>Representing an input sample at time t, f (x) representing a nonlinear activation function in the RNN;
(2) X n is subjected to forward mapping for L times to obtain characteristic output The feature output of the RNN is subjected to forward mapping of a modal identification network to obtain/>Is marked as/>K is the number of decoders; wherein h m is the node number of the hidden layer of the modal identification network; w m、bm represents the weights and biases, respectively, of hidden layer features mapping inputs to the modal identification network,/> W p、bp represents the weights and biases, respectively, that map hidden layer features of the modality identification network to the output of the modality identification network,/>
(3) The feature output of RNN is mapped forward by the ith decoder to obtainWherein σ is a nonlinear activation function; /(I)Weights and biases representing hidden layer features mapping inputs to the ith decoder, respectively,/>Weights and offsets respectively representing mapping of hidden layer features of the ith decoder to outputs of the ith decoder,/>H d is the node number of the hidden layer of the decoder network;
During training, the loss function is defined as:
Where N is the number of sequences used to train the model, alpha and beta are adjustable hyper-parameters, And L entr2 is information entropy output through a modal identification network and is used for obtaining more accurate modal identification precision, and meanwhile, the model is prevented from collapsing to a single mode in the training process and falling into a local optimal solution, and the calculation method is as shown in the following formulas:
further, the calculation mode of the S4 detection index WSPE is as follows:
the detection index WSPE in S5 o
Further, the loss function is minimized by the gradient descent method in S3.
The beneficial effects of the invention are as follows:
(1) The invention relates to a process detection model, which aims at the characteristics of strong dynamic property and nonlinearity of a one-section furnace process, wherein an encoder part consists of a group of cyclic neural networks, and the last hidden layer output of the encoder part is used for reconstructing a sample at the last moment of a sequence through a decoder. In addition, since a multi-mode condition exists in a one-stage furnace process, in order to prevent confusion among features of a plurality of modes, a decoder part is composed of a plurality of sub-decoder parts, and weights output by a mode identification network are weighted to obtain a final reconstruction value. The network parameters are trained by adding the weight reconstruction errors, and the information entropy aiming at the weight is added into the loss function, so that a more accurate mode identification effect is obtained, and the network is prevented from collapsing towards a single mode. Finally, based on the designed neural network model, a weighted square prediction error is constructed to indicate whether a fault occurs in the process, and fault variables are identified through contribution.
(2) Because the method is based on deep learning, compared with the traditional process monitoring model, the method has stronger process complex feature mining capability. In addition, due to the addition of the RNN structure, the model can effectively extract dynamic information in the process, so that the mode identification is more accurate, and meanwhile, the mined dynamic characteristics can well assist various tasks in process monitoring, so that the monitoring accuracy is improved. Compared with the traditional single-model-based process monitoring model, the method has the advantages that a plurality of decoders are added in the model structural design, so that feature extraction aiming at each mode is more accurate, and further performance is improved in a multi-mode data-oriented process monitoring task.
Drawings
FIG. 1 is a flow chart of a method of performing a section of furnace process monitoring based on a mixed sequence network;
fig. 2 is a schematic diagram of a model structure of a mixed sequence network.
Detailed Description
The objects and effects of the present invention will become more apparent from the following detailed description of the preferred embodiments and the accompanying drawings, it being understood that the specific embodiments described herein are merely illustrative of the invention and not limiting thereof.
The process monitoring model based on the mixed sequence network changes the encoder into a cyclic neural network based on the original self-encoder so as to mine dynamic characteristics among data, and in addition, a decoder part consists of a plurality of sub-decoders in consideration of multi-modal characteristics in a one-stage furnace process, and simultaneously, a modal identification network output modal identification weight is constructed for subsequent weighted reconstruction.
As shown in fig. 1, the method for monitoring a one-stage furnace process based on a mixed sequence network mainly comprises two parts of offline modeling and online monitoring, wherein the online monitoring part comprises fault detection and fault identification aiming at a fault sample; the method comprises the following specific steps:
S1: constructing a process monitoring model based on a mixed sequence network, and performing feature mining on a primary furnace process;
As shown in fig. 2, the process monitoring model includes three parts, namely an encoder, a modal identification network and a decoder; the encoder is RNN and is used for dynamic characteristics in the process of mining; the mode identification network outputs a group of weights through a softmax layer after passing through a hidden layer after being connected with the hidden layer output at the last moment of the encoder, and the weights are used for indicating the mode of the current sequence sample; the decoder is also connected with the hidden layer output at the last moment of the encoder and is used for reconstructing the sample at the last moment in the input sequence; the decoder consists of a plurality of single-layer neural networks, and the final reconstruction value is obtained after each neural network output is weighted and summed through the weight output by the modal identification network;
S2: collecting process data under normal working conditions of a section of furnace process, constructing a data set, setting a sequence length L at the same time, serializing the data set, and taking the obtained sequence data set as a training data set X for model training; wherein the nth input sequence is
S3: inputting a training data set X into a process monitoring model based on a mixed sequence network, carrying out forward propagation, carrying out weighted summation after obtaining modal identification network output and each decoder output to obtain a reconstruction value, taking a reconstruction error as a loss function, minimizing the loss function through an iterative method (carrying out iterative updating on model parameters by using optimization methods such as gradient descent) until the model parameters are converged or the maximum iterative times are reached, and obtaining a trained process monitoring model; the method specifically comprises the following substeps:
(1) X n is brought into a process monitoring model constructed by S1, and X n is transmitted forward to obtain hidden layer output of RNN
Wherein U e denotes the weights of hidden layer features mapping inputs to the RNN,M is the variable number of the input samples; w e denotes the weight to map the hidden layer output at time t-1 in the RNN to the hidden layer output at time t in the RNN,H e is the node number of the RNN hidden layer; /(I)And/>Hidden layer output representing time t and time t-1 respectively,/>Representing an input sample at the time t, f (X) represents a nonlinear activation function in the RNN, and sequentially bringing each element of X n into the mapping for iteration, so that a final RNN characteristic output can be obtained;
(2) X n is subjected to forward mapping for L times to obtain the characteristic output of RNN The feature output of the RNN is subjected to forward mapping of a modal identification network to obtain/>For indicating modal information, noted asK is the number of decoders, wherein the size of each element indicates the probability size of the current sequence sample belonging to a certain mode; wherein h m is the node number of the hidden layer of the modal identification network; w m、bm represents the weights and biases, respectively, of hidden layer features mapping inputs to the modal identification network,/>W p、bp represents the weights and biases, respectively, that map hidden layer features of the modality identification network to the output of the modality identification network,/>
(3) The feature output of RNN is mapped forward by the ith decoder to obtain the reconstruction value of the ith decoder, and the mapping function is thatWherein σ is a nonlinear activation function; /(I)Weights and biases representing hidden layer features mapping inputs to the ith decoder, respectively,/>Weights and offsets respectively representing mapping of hidden layer features of the ith decoder to outputs of the ith decoder,/> H d is the node number of the hidden layer of the decoder network;
During training, the loss function is defined as:
Where N is the number of sequences used to train the model, alpha and beta are adjustable hyper-parameters, And L entr2 is information entropy output through a modal identification network and is used for obtaining more accurate modal identification precision, and meanwhile, the model is prevented from collapsing to a single mode in the training process and falling into a local optimal solution, and the calculation method is as shown in the following formulas:
S4: and (3) indicating whether the one-stage furnace process fails or not by establishing a detection index based on the weighted square prediction error. I.e., using the training data to calculate the detection index WSPE, After WSPE sets of all training data are obtained, calculating a control limit con wspe by using a kernel density estimation method;
S5: constructing an input sequence of a required length L by utilizing a section of furnace on-line detection sample x and samples at the previous L-1 moments, substituting the input sequence into a trained process monitoring model in S4 to obtain reconstruction output of a decoder of x and output p= [ p 1 p2 … pK ] of a modal identification network; the reconstructed output of the ith sub-decoder is noted as
S6: the detection index WSPE o is calculated using the on-line samples and their reconstructed outputs,Comparing the detection index with a control limit con wspe, wherein when the sample is normal, the model can reconstruct input well, so when WSPE o≤conwspe, the online sample is a normal sample; when WSPE o＞conwspe, the current sample is considered as a fault sample, and the fault sample is further subjected to fault identification, namely a fault variable is identified; and respectively calculating the contribution degree of the corresponding variable to the fault index by using the online sample and the reconstruction value output by each decoder, and then weighting each result by using the weight output by the modal identification network to obtain the final variable contribution degree so as to indicate the possibility of the fault of the variable. Let x= [ x 1 x2 … xm ] be the on-line sample, and the reconstruction value of the ith decoder after model is/>The calculation method of the contribution index of the j-th variable is as follows:
s7: and regarding the variable with higher contribution degree as a fault variable according to the requirement.
In order to detect the fault detection effect of the process monitoring model based on the mixed sequence network, the off-line data are utilized to respectively calculate the detection rate (FDR) and False Alarm Rate (FAR) of the model
Where N f is the number of failure samples, N n is the number of normal samples, and N fa and N na are the number of alarm samples in the failure and normal samples, respectively.
The advantages of the method according to the invention are described below in connection with practical process experiments of a one-stage furnace process. The data in the experiment are acquired on site, and normal process data under two modes are extracted for training of the model by investigating operation logs and observing data distribution, and a group of fault data sets are selected for fault detection experiments under each mode. The comparison method comprises a traditional multi-mode fault detection method GMM-PCA and a deep learning-based method SAE, MAE. Parameters of all models in the experiment were determined by achieving less than 5% FAR in the normal dataset. The fault detection rates of the methods are shown in the following table.
Table 1 one stage furnace process fault detection results
As can be seen from the results shown in table 1, the mixed sequence network has a significant performance improvement compared with the other three methods, and the fault detection rate in each mode is improved by more than 10%.
It will be appreciated by persons skilled in the art that the foregoing description is a preferred embodiment of the invention, and is not intended to limit the invention, but rather to limit the invention to the specific embodiments described, and that modifications may be made to the technical solutions described in the foregoing embodiments, or equivalents may be substituted for elements thereof, for the purposes of those skilled in the art. Modifications, equivalents, and alternatives falling within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (4)
1. A method for monitoring a furnace process based on a mixed sequence network, the method comprising the steps of:
S1: constructing a process monitoring model based on a mixed sequence network, and performing feature mining on a primary furnace process;
the process monitoring model comprises an encoder, a modal identification network and a decoder; the encoder is RNN and is used for dynamic characteristics in the process of mining; the mode identification network outputs a group of weights through a softmax layer after passing through a hidden layer after being connected with the hidden layer output at the last moment of the encoder, and the weights are used for indicating the mode of the current sequence sample; the decoder is also connected with the hidden layer output at the last moment of the encoder and is used for reconstructing the sample at the last moment in the input sequence; the decoder consists of a plurality of single-layer neural networks, and the final reconstruction value is obtained after each neural network output is weighted and summed through the weight output by the modal identification network;
s2: collecting process data under normal working conditions of a section of furnace process, constructing a data set, setting a sequence length L at the same time, serializing the data set, and taking the obtained sequence data set as a training data set X for model training; wherein the nth input sequence is
S3: inputting the training data set X into a process monitoring model based on a mixed sequence network, performing forward propagation to obtain a reconstruction value, and minimizing a loss function by an iterative method until model parameters converge or the maximum iterative times are reached to obtain a trained process monitoring model;
S4: calculating a detection index WSPE by using training data, and calculating a control limit con wspe by using a nuclear density estimation method;
S5: constructing an input sequence of a required length L by utilizing a section of furnace on-line detection sample x and samples at the previous L-1 moments, substituting the input sequence into a trained process monitoring model in S3 to obtain reconstruction output of a decoder of x and output p= [ p 1 p2…pK ] of a modal identification network; the reconstructed output of the ith sub-decoder is noted as
S6: calculating a detection index WSPE o by using the online sample and the reconstruction output thereof, and comparing the detection index with a control limit con wspe, wherein the online sample is a normal sample when WSPE o≤conwspe; when WSPE o＞conwspe, the current sample is considered as a fault sample, and the fault sample is further subjected to fault identification; let x= [ x 1 x2…xm ] be the line sample, the reconstruction value of the ith decoder after the model is taken asThe calculation method of the contribution index of the j-th variable is as follows:
s7: and regarding the variable with higher contribution degree as a fault variable according to the requirement.
2. The method for one-stage furnace process monitoring based on a mixed sequence network according to claim 1, wherein S3 is implemented by the sub-steps of:
(1) X n is brought into a process monitoring model constructed by S1, and X n is transmitted forward to obtain hidden layer output of RNN
Wherein U e denotes the weights of hidden layer features mapping inputs to the RNN,M is the variable number of the input samples; w e denotes the weight to map the hidden layer output at time t-1 in the RNN to the hidden layer output at time t in the RNN,H e is the node number of the RNN hidden layer; /(I)And/>Hidden layer output representing time t and time t-1 respectively,/>Representing an input sample at time t, f (x) representing a nonlinear activation function in the RNN;
(2) X n is subjected to forward mapping for L times to obtain characteristic output The feature output of the RNN is subjected to forward mapping of a modal identification network to obtain/>Is marked as/>K is the number of decoders; wherein h m is the node number of the hidden layer of the modal identification network; w m、bm represents the weights and biases, respectively, of hidden layer features mapping inputs to the modal identification network,/> W p、bp represents the weights and biases, respectively, that map hidden layer features of the modality identification network to the output of the modality identification network,/>
(3) The feature output of RNN is mapped forward by the ith decoder to obtainWherein σ is a nonlinear activation function; /(I)Weights and biases representing hidden layer features mapping inputs to the ith decoder, respectively,/> Weights and offsets respectively representing mapping of hidden layer features of the ith decoder to outputs of the ith decoder,/>H d is the node number of the hidden layer of the decoder network;
During training, the loss function is defined as:
Where N is the number of sequences used to train the model, alpha and beta are adjustable hyper-parameters, And L entr2 is information entropy output through a modal identification network and is used for obtaining more accurate modal identification precision, and meanwhile, the model is prevented from collapsing to a single mode in the training process and falling into a local optimal solution, and the calculation method is as shown in the following formulas:
3. The method for monitoring a furnace process based on a mixed sequence network according to claim 1, wherein the S4 detection index WSPE is calculated as follows:
the detection index WSPE in S5 o
4. The method for one-stage furnace process monitoring based on a mixed sequence network according to claim 1, wherein the loss function is minimized in S3 by a gradient descent method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111348802.8A CN114021469B (en) | 2021-11-15 | 2021-11-15 | Method for monitoring one-stage furnace process based on mixed sequence network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111348802.8A CN114021469B (en) | 2021-11-15 | 2021-11-15 | Method for monitoring one-stage furnace process based on mixed sequence network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114021469A CN114021469A (en) | 2022-02-08 |
| CN114021469B true CN114021469B (en) | 2024-06-21 |
Family
ID=80064227
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111348802.8A Active CN114021469B (en) | 2021-11-15 | 2021-11-15 | Method for monitoring one-stage furnace process based on mixed sequence network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114021469B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114854920B (en) * | 2022-05-06 | 2023-06-30 | 浙江大学 | GRU self-encoder blast furnace abnormity monitoring method embedded by Gaussian mixture model |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107798425A (en) * | 2017-10-16 | 2018-03-13 | 中国科学院地理科学与资源研究所 | A kind of space-time based on big data obscures degrees of exposure assessment system and method |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150277416A1 (en) * | 2014-03-31 | 2015-10-01 | Mitsubishi Electric Research Laboratories, Inc. | Method for Anomaly Detection in Discrete Manufacturing Processes |
| US11205121B2 (en) * | 2018-06-20 | 2021-12-21 | Disney Enterprises, Inc. | Efficient encoding and decoding sequences using variational autoencoders |
| CN109800875A (en) * | 2019-01-08 | 2019-05-24 | 华南理工大学 | Chemical industry fault detection method based on particle group optimizing and noise reduction sparse coding machine |
-
2021
- 2021-11-15 CN CN202111348802.8A patent/CN114021469B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107798425A (en) * | 2017-10-16 | 2018-03-13 | 中国科学院地理科学与资源研究所 | A kind of space-time based on big data obscures degrees of exposure assessment system and method |
Non-Patent Citations (1)
| Title |
|---|
| 基于循环神经网络的半监督动态软测量建模方法;邵伟明;葛志强;李浩;宋执环;;电子测量与仪器学报;20191115(第11期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114021469A (en) | 2022-02-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108875771B (en) | Fault classification model and method based on sparse Gaussian Bernoulli limited Boltzmann machine and recurrent neural network | |
| CN113723010B (en) | Bridge damage early warning method based on LSTM temperature-displacement correlation model | |
| CN113011085B (en) | Equipment digital twin modeling method and system | |
| CN111046961B (en) | Fault classification method based on bidirectional long-time and short-time memory unit and capsule network | |
| CN111079836A (en) | Process data fault classification method based on pseudo label method and weak supervised learning | |
| CN115758290A (en) | Fan gearbox high-speed shaft temperature trend early warning method based on LSTM | |
| CN112147432A (en) | BiLSTM module based on attention mechanism, transformer state diagnosis method and system | |
| CN113836783B (en) | Digital regression model modeling method for main beam temperature-induced deflection monitoring reference value of cable-stayed bridge | |
| CN115510975A (en) | Multivariable time sequence abnormality detection method and system based on parallel Transomer-GRU | |
| CN114169091A (en) | Method for establishing prediction model of residual life of engineering mechanical part and prediction method | |
| CN114021469B (en) | Method for monitoring one-stage furnace process based on mixed sequence network | |
| CN111598187A (en) | Progressive integrated classification method based on kernel width learning system | |
| WO2023231374A1 (en) | Semi-supervised fault detection and analysis method and apparatus for mechanical device, terminal, and medium | |
| CN113988210A (en) | Method and device for restoring distorted data of structure monitoring sensor network and storage medium | |
| CN115841004A (en) | Strip steel hot rolling process mechanical property soft measurement method and device based on multidimensional data | |
| CN112731890A (en) | Power plant equipment fault detection method and device | |
| CN111428788A (en) | Deep learning-based multi-fault diagnosis method and system for steam turbine generator set rotor | |
| CN115345222A (en) | Fault classification method based on TimeGAN model | |
| CN115438897A (en) | Industrial process product quality prediction method based on BLSTM neural network | |
| CN114239397A (en) | Soft measurement modeling method based on dynamic feature extraction and local weighted deep learning | |
| CN116894180B (en) | Product manufacturing quality prediction method based on different composition attention network | |
| CN116628444A (en) | Water quality early warning method based on improved meta-learning | |
| CN116432359A (en) | Variable topology network tide calculation method based on meta transfer learning | |
| CN114897047B (en) | Multi-sensor data drift detection method based on depth dictionary | |
| CN116304587A (en) | Rolling bearing degradation trend prediction method based on CAE and AGRU |
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 |