CN114854920B - GRU self-encoder blast furnace abnormity monitoring method embedded by Gaussian mixture model - Google Patents
GRU self-encoder blast furnace abnormity monitoring method embedded by Gaussian mixture model Download PDFInfo
- Publication number
- CN114854920B CN114854920B CN202210486776.3A CN202210486776A CN114854920B CN 114854920 B CN114854920 B CN 114854920B CN 202210486776 A CN202210486776 A CN 202210486776A CN 114854920 B CN114854920 B CN 114854920B
- Authority
- CN
- China
- Prior art keywords
- encoder
- spe
- gaussian
- gru
- mixture 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
Images
Classifications
-
- C—CHEMISTRY; METALLURGY
- C21—METALLURGY OF IRON
- C21B—MANUFACTURE OF IRON OR STEEL
- C21B7/00—Blast furnaces
- C21B7/24—Test rods or other checking devices
-
- C—CHEMISTRY; METALLURGY
- C21—METALLURGY OF IRON
- C21B—MANUFACTURE OF IRON OR STEEL
- C21B5/00—Making pig-iron in the blast furnace
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- C—CHEMISTRY; METALLURGY
- C21—METALLURGY OF IRON
- C21B—MANUFACTURE OF IRON OR STEEL
- C21B2300/00—Process aspects
- C21B2300/04—Modeling of the process, e.g. for control purposes; CII
-
- 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
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The invention discloses a blast furnace anomaly monitoring method of a GRU self-encoder embedded by a Gaussian mixture model, which uses GRU as dynamic characteristics of encoder capture data; then utilizing a Gaussian mixture model to cluster data in a hidden variable space, and adapting to different working modes of the blast furnace; the hidden variables are reconstructed by a decoder to obtain reconstruction errors, SPE statistics and the upper limit thereof based on probability are calculated, the monitoring purpose is realized, and the correlation degree of different variables and faults is calculated according to the reconstruction error increment when the faults occur; the dynamic characteristics of capturing data using the GRU as an encoder are: the self-encoder is divided into an encoder and a decoder, and GRU is used as the encoder part of the self-encoder. The invention effectively reduces the false alarm rate of the algorithm, can timely early warn process faults and provides fault related variable information.
Description
Technical Field
The invention belongs to the field of industrial process monitoring, and particularly relates to a GRU self-encoder blast furnace abnormality monitoring method embedded by a Gaussian mixture model.
Technical Field
The steel industry is an important pillar industry of modern industry and national economy in China, and is also an indispensible important component part of national economy. The blast furnace ironmaking is a core link of the whole iron and steel industry, and is also a key process of energy mass flow conversion in the iron and steel manufacturing process, and the energy consumption of the blast furnace ironmaking accounts for 70% of the total energy consumption of iron and steel production. The blast furnace ironmaking process safe operation is the important of deep energy conservation, emission reduction, quality improvement and synergy in the iron and steel industry, and an accurate abnormal monitoring model can timely provide early warning for operators, adjust the furnace condition in advance and avoid dangerous situations.
In the blast furnace ironmaking process, the blast furnace condition is abnormal due to the reasons of fluctuation of raw material quality, misoperation of personnel, equipment abnormality and the like. In case of abnormal blast furnace conditions, the problems of increased fuel ratio, increased downtime overhaul time, substandard molten iron quality and the like are caused, so that not only are the important losses of resources and equipment caused, but also the service life of the blast furnace is reduced, and even casualties are possibly caused by accidents. Therefore, the safety of the blast furnace ironmaking process is improved, particularly the frequency of blast furnace accidents and abnormal furnace conditions is reduced, and the method has very important significance for reducing energy consumption, ensuring the safety of equipment and personnel and improving the economic benefit of the steel production process. However, according to the investigation result, the current fault recognition mainly depends on the working experience of operators, which brings about two disadvantages: firstly, accumulation of fault experience takes years, and secondly, continuous observation of data curves by operators is a boring task, and fatigue errors are easy to cause. With the increasing level of industrial informatization, data driven process monitoring has become a key technology to improve process industry safety, quality and operational efficiency. The large-scale blast furnace data also has the characteristics of non-Gaussian distribution, time variation and the like, and the abnormality monitoring and diagnosis become a leading-edge hot spot and difficulty of the current world metallurgical technology research.
Disclosure of Invention
Aiming at the existing problems of blast furnace anomaly monitoring, the invention provides a GRU self-encoder blast furnace anomaly monitoring method embedded by a Gaussian mixture model, which adapts to the dynamic property of blast furnace data by utilizing the dynamic characteristic of a GRU capturing process; and a Gaussian mixture model is added on the basis, so that the method is suitable for different working modes of the blast furnace, and the final model can realize rapid and accurate early warning of abnormal furnace conditions of the blast furnace and provide relevant information of fault variables.
A GRU self-encoder blast furnace abnormality monitoring method embedded by a Gaussian mixture model utilizes GRU as an encoder to capture dynamic characteristics of data; then utilizing a Gaussian mixture model to cluster data in a hidden variable space, and adapting to different working modes of the blast furnace; the hidden variables are reconstructed by a decoder to obtain reconstruction errors, SPE statistics and the upper limit thereof based on probability are calculated, the monitoring purpose is realized, and the correlation degree of different variables and faults is calculated according to the reconstruction error increment when the faults occur;
the dynamic characteristics of capturing data using the GRU as an encoder are:
the self-encoder is divided into an encoder and a decoder, and GRU is used as the encoder part of the self-encoder.
The model utilizes a Gaussian mixture model to cluster data in a hidden variable space as follows:
inputting a process variable X (n multiplied by m) into an encoder, namely a GRU model, encoding to obtain a hidden variable Z (n multiplied by a) with a dimension of a (a < m), performing soft clustering in a hidden variable space by using a Gaussian mixture model, setting the clustering number k according to the number of working mode types of the process, and solving through an EM algorithm to obtain weights, mean values and variances of different Gaussian components: Ω= { (pi) 1 ,μ 1 ,Σ 1 ),...,(π k ,μ k ,Σ k ) Of which (pi) i ,μ i ,Σ i ) The weights, means and variances of the ith gaussian component are represented in turn.
The decoder is a single-layer full-connection layer.
The probability-based SPE statistics are: after the process variable X (n X m) has been reconstructed from the encoder, the output is reconstructed from the input variable X
Calculating reconstruction error e and SPE statistics:
SPE＝e T e
in the offline training stage, the Gaussian mixture model clusters different samples into k Gaussian components, and then calculates the upper limit SPE of the monitoring statistics of the SPE statistics of the samples in each Gaussian component by using a kernel density estimation method UCL The method comprises the steps of carrying out a first treatment on the surface of the The on-line monitoring stage is performed on a new sample x new After arrival, the probability of belonging to different Gaussian components is calculated
Then dynamically calculating monitoring statistic upper limit SPE by probability UCLnew ：
Wherein the method comprises the steps of
Probability of belonging to the jth gaussian component for a new sample.
The correlation degree of different variables and faults is calculated according to the reconstruction error increment when the faults occur: assuming that the moment when the fault occurs is t moment, the reconstruction error of the sample is marked as e t Then the reconstruction error increment is |e t -e t-1 And judging the correlation between the error and the fault according to the reconstruction error increment of each variable at the moment of occurrence of the fault.
The method comprises the following specific processes:
step 1: offline modeling;
step 1.1: collecting the process variable X (n multiplied by m) of the blast furnace sensor, and recording the normalized data as X train (n x m), the number of samples is n, the number of variables is m, and the number of samples is input into a GRU encoder for feature extraction to obtain hidden variables Z (n x a);
step 1.2: determining the number of Gaussian components of the Gaussian mixture model as k according to the number of the working mode varieties of the process, solving the hidden variable Z by using the Gaussian mixture model to obtain the weight, the mean value and the covariance omega = { (pi) of each Gaussian component 1 ,μ 1 ,Σ 1 ),...,(π k ,μ k ,Σ k )}；
Step 1.3: decoding the hidden variable Z by a decoder formed by a single-layer full-connection layer to obtain a decoding result
Calculating reconstruction error e and SPE statistics of samples, and then sequentially obtaining monitoring statistics upper limit SPE of different Gaussian components by using a kernel density estimation method according to clustering results of the samples UCL ；
Step 2: on-line monitoring;
step 2.1: with new sample x of blast furnace new When arriving, the data is normalized, and the normalized data is recorded as x test Then input it into the self-training parameters in offline modelingEncoder for obtaining reconstruction result
And clustering results of the Gaussian mixture model, and calculating a reconstruction error e test And calculating the monitoring statistic upper limit SPE according to the probability of the clustering result UCLtest ：
SPE test ＝e test T e test
Wherein the method comprises the steps of
Probability of belonging to the jth gaussian component for the new sample, while SPE UCLj The upper limit of the monitoring statistics for the jth gaussian component.
Comparison statistics SPE test And statistics upper limit SPE UCLtest If it is SPE test ≥SPE UCLtest And if so, representing that the fault occurs, otherwise, the device is in a normal operation state.
If the fault is judged to occur at the moment t, calculating the reconstruction error increment |e of the moment t t -e t-1 And judging the magnitude of the correlation between each variable and the fault according to the magnitude.
The invention has the beneficial effects that:
1. the structure of the self-encoder is modified, and dynamic information of the process is effectively extracted by using the GRU.
2. The embedding of the Gaussian mixture model effectively adapts to different working modes of the process.
3. The model considers the dynamic property and non-Gaussian property (different working modes) of the process, effectively reduces the false alarm rate of the algorithm, can timely early warn the process faults and provides fault related variable information.
Drawings
FIGS. 1-1,1-2,1-3,1-4 are graphs of the trend of the 4 key variables (hot air pressure, full pressure differential, theoretical combustion temperature, kinetic energy of blowing) respectively.
Fig. 2 is an autocorrelation analysis of 4 key variables.
FIGS. 3-1 and 3-2 are two-dimensional distribution diagrams of 4 key variables.
Fig. 4 is a schematic diagram of a monitoring system.
FIG. 5 is a graph of SPE statistic monitoring.
Fig. 6 is a variable reconstruction error delta map at the time of occurrence of a failure.
Detailed Description
The invention is further illustrated in the following figures and examples.
A GRU self-encoder blast furnace abnormality monitoring method embedded by a Gaussian mixture model utilizes GRU as an encoder to capture dynamic characteristics of data; soft clustering data in hidden variable space by utilizing a Gaussian mixture model, and adapting to different working modes of the blast furnace; the hidden variable is reconstructed by a decoder to obtain a reconstruction error, SPE statistics and an upper limit thereof based on probability are calculated, the monitoring purpose is realized, and the correlation degree of different variables and faults is calculated according to the reconstruction error increment when the faults occur.
The dynamic characteristics of capturing data using the GRU as an encoder are: the self-encoder is divided into an encoder and a decoder, the GRU is used as the encoder part of the self-encoder, the time step of the GRU determines that the input contains variables of a plurality of moments, and the selection of the parameters can be determined by using a cross-validation method.
The model utilizes a Gaussian mixture model to cluster data in a hidden variable space as follows:
inputting a process variable X (n multiplied by m) into an encoder, namely a GRU model, encoding to obtain an hidden variable Z (n multiplied by a) with a dimension of a (a < m), clustering in a hidden variable space by using a Gaussian mixture model, setting the clustering number k according to the number of working mode types of the process, and solving through an EM algorithm to obtain weights, mean values and variances of different Gaussian components: Ω= { (pi) 1 ,μ 1 ,Σ 1 ),...,(π k ,μ k ,Σ k ) Of which (pi) i ,μ i ,Σ i ) Sequentially representing the weight of the ith Gaussian component,Mean and variance.
The decoder is a single-layer full-connection layer.
The probability-based SPE statistics are:
after the process variable X (n X m) has been reconstructed from the encoder, the output is reconstructed from the input variable X
Calculating reconstruction error e and SPE statistics:
SPE＝e T e
in the offline training stage, the Gaussian mixture model clusters different samples into k Gaussian components, and then calculates the upper limit SPE of the monitoring statistics of the SPE statistics of the samples in each Gaussian component by using a kernel density estimation method UCL The method comprises the steps of carrying out a first treatment on the surface of the The on-line monitoring stage is performed on a new sample x new After arrival, calculate the probability of belonging to different Gaussian components
Then dynamically calculating monitoring statistic upper limit SPE by probability UCLnew ：
Wherein the method comprises the steps of
Probability of belonging to the jth gaussian component for a new sample.
The correlation degree of different variables and faults is calculated according to the reconstruction error increment when the faults occur:
assuming that the moment when the fault occurs is t moment, the reconstruction error of the sample is marked as e t Then the reconstruction error increment is |e t -e t-1 I, each of the occurrence timings of the faults can be based onThe reconstruction error increment of the variable judges the correlation between the variable and the fault.
The specific flow is as follows:
step 1: offline modeling;
step 1.1: collecting the process variable X (n multiplied by m) of the blast furnace sensor, and recording the normalized data as X train (n x m), the number of samples is n, the number of variables is m, and the number of samples is input into a GRU encoder for feature extraction to obtain hidden variables Z (n x a);
step 1.2: determining the number of Gaussian components of the Gaussian mixture model as k according to the number of the working mode varieties of the process, solving the hidden variable Z by using the Gaussian mixture model to obtain the weight, the mean value and the covariance omega = { (pi) of each Gaussian component 1 ,μ 1 ,Σ 1 ),...,(π k ,μ k ,Σ k )}；
Step 1.3: decoding the hidden variable Z by a decoder formed by a single-layer full-connection layer to obtain a decoding result
Calculating reconstruction error e and SPE statistics of samples, and then sequentially obtaining monitoring statistics upper limit SPE of different Gaussian components by using a kernel density estimation method according to clustering results of the samples UCL 。
Step 2: on-line monitoring;
step 2.1: with new sample x of blast furnace new When arriving, the data is normalized, and the normalized data is recorded as x test Then inputting the obtained information into a self-encoder with trained parameters in offline modeling to obtain a reconstruction result
And clustering results of the Gaussian mixture model, and calculating a reconstruction error e test And calculating the monitoring statistic upper limit SPE according to the probability of the clustering result UCLtest ：
SPE test ＝e test T e test
Wherein the method comprises the steps of
Probability of belonging to the jth gaussian component for the new sample, while SPE UCLj The upper limit of the monitoring statistics for the jth gaussian component.
Comparison statistics SPE test And statistics upper limit SPE UCLtest If it is SPE test ≥SPE UCLtest And if so, representing that the fault occurs, otherwise, the device is in a normal operation state.
If the fault is judged to occur at the moment t, calculating the reconstruction error increment |e of the moment t t -e t-1 And judging the magnitude of the correlation between each variable and the fault according to the magnitude.
Examples
1. Introduction of blast furnace process variables
This experiment was directed to a 2650 cubic meter blast furnace from Guangxi Liuzhou Steel group. The furnace data were sampled once for 10 seconds on average and contained 27 variables as shown in table 1.
Table 1 variable list of datasets
There are repeated variables in the table, such as multiple pressures, temperatures, etc., because the sensors are placed in different locations or are redundant, and the repeated variables are averaged in a subsequent process to avoid repetition of the variables.
According to the experience of operators, the hot air pressure, the total pressure drop, the theoretical combustion temperature and the blast kinetic energy are key indexes for reflecting the running state of the blast furnace. The variation of these four key indicators throughout the day is thus selected as shown in figures 1-1,1-2,1-3, 1-4.
It can be seen from fig. 1-1,1-2,1-3,1-4 that there is a peak-like disturbance of some variables at intervals, which is caused by switching of the stoves of the blast furnace body, the blast furnace often has more than one stove, when one stove is blowing air to the blast furnace, the other stoves need to store heat and then switch after a period of time to ensure uninterrupted operation of the blast furnace, which results in two significant modes of operation of the blast furnace, namely normal operation and stove switching modes.
2. Blast furnace process variable analysis
Dynamic properties:
the autocorrelation analysis of 4 representative variables is shown in fig. 2. The dynamics of the data refer to the correlation between the variables at different times. It can be seen from the figure that the autocorrelation coefficients of 4 typical variables are high, and the dynamics of the variables are obvious. Dynamic considerations are therefore necessary for process monitoring.
Non-gaussian distribution:
because the blast furnace has different working modes and transitional states, the data can be in a non-Gaussian distribution characteristic. The 4 key variable distribution characteristics are shown in fig. 3-1 and 3-2, and the visible data of the graph show remarkable non-gaussian distribution conditions.
3. Algorithm framework
The framework of the overall algorithm is shown in fig. 4.
4. Complete flow of GRU self-encoder blast furnace anomaly monitoring method embedded by Gaussian mixture model
Step 1: offline modeling;
step 1.1: collecting the process variable X (n multiplied by m) of the blast furnace sensor, and recording the normalized data as X train (n x m), the number of samples is n, the number of variables is m, and the number of samples is input into a GRU encoder for feature extraction to obtain hidden variables Z (n x a);
step 1.2: determining the number of Gaussian components of the Gaussian mixture model as k according to the number of the working mode varieties of the process, solving the hidden variable Z by using the Gaussian mixture model to obtain the weight, the mean value and the covariance omega = { (pi) of each Gaussian component 1 ,μ 1 ,Σ 1 ),...,(π k ,μ k ,Σ k )}；
Step 1.3: decoding the hidden variable Z by a decoder formed by a single-layer full-connection layer to obtainDecoding result
Calculating reconstruction error e and SPE statistics of samples, and then sequentially obtaining monitoring statistics upper limit SPE of different Gaussian components by using a kernel density estimation method according to clustering results of the samples UCL ；
Step 2: on-line monitoring;
step 2.1: with new sample x of blast furnace new When arriving, the data is normalized, and the normalized data is recorded as x test Then inputting the obtained information into a self-encoder with trained parameters in offline modeling to obtain a reconstruction result
And clustering results of the Gaussian mixture model, and calculating a reconstruction error e test And calculating the monitoring statistic upper limit SPE according to the probability of the clustering result UCLtest ：
SPE test ＝e test T e test
Wherein the method comprises the steps of
Probability of belonging to the jth gaussian component for the new sample, while SPE UCLj The upper limit of the monitoring statistics for the jth gaussian component.
Comparison statistics SPE test And statistics upper limit SPE UCLtest If it is SPE test ≥SPE UCLtest And if so, representing that the fault occurs, otherwise, the device is in a normal operation state.
If the fault is judged to occur at the moment t, calculating the reconstruction error increment |e of the moment t t -e t-1 And judging the magnitude of the correlation between each variable and the fault according to the magnitude.
5. And finally outputting a result by the system:
taking 100000 pieces of historical data of the blast furnace sensor as a training set, 30000 pieces of historical data as a cross verification set and finally 20500 pieces of historical data as a test set to verify the system, wherein the test set of 20500 samples comprises a fault of abnormal blast furnace temperature, and the fault time of a worker mark is 20435.
The effect of the test set after model training is completed is shown in fig. 5 (the last 5000 samples are chosen for ease of presentation). The small circles in the figure represent false positive sample points, while the upward arrows represent that a fault has actually occurred. The moment of detecting the fault by the monitoring system is 20380, which is 55 sampling points earlier than the moment marked by workers, namely, the fault is early-warned 550 seconds earlier, and the effectiveness of the system is fully reflected. And finally, the system provides error increment of reconstruction of the variable at the moment of the fault, as shown in fig. 6, the variable related to the fault can be judged according to the reconstruction error. The sequence of the 20 variables is as follows: oxygen enrichment rate, air permeability index, standard wind speed, oxygen enrichment flow, cold wind flow, blast kinetic energy, stove belly gas quantity, stove belly gas index, theoretical combustion temperature, full pressure difference, actual wind speed, cold wind temperature, hot wind temperature, resistance coefficient, set coal injection quantity, actual coal injection quantity in this hour, top temperature, top pressure, cold wind pressure, hot wind pressure.
The embodiments in the foregoing description may be further combined or replaced, and the embodiments are merely illustrative of the preferred embodiments of the present invention and are not intended to limit the spirit and scope of the present invention, and various changes and modifications made by those skilled in the art to which the present invention pertains without departing from the spirit of the present invention. The scope of the invention is given by the appended claims and any equivalents thereof.
Claims (5)
1. A GRU self-encoder blast furnace abnormality monitoring method embedded by a Gaussian mixture model is characterized in that GRU is used as an encoder to capture dynamic characteristics of data; then utilizing a Gaussian mixture model to cluster data in a hidden variable space, and adapting to different working modes of the blast furnace; the hidden variables are reconstructed by a decoder to obtain reconstruction errors, SPE statistics and the upper limit thereof based on probability are calculated, the monitoring purpose is realized, and the correlation degree of different variables and faults is calculated according to the reconstruction error increment when the faults occur;
the dynamic characteristics of capturing data using the GRU as an encoder are:
the self-encoder is divided into an encoder and a decoder, and GRU is used as the encoder part of the self-encoder;
the method comprises the following steps:
step 1: offline modeling;
step 1.1: collecting the process variable X (n multiplied by m) of the blast furnace sensor, and recording the normalized data as X train (n x m), the number of samples is n, the number of variables is m, and the number of samples is input into a GRU encoder for feature extraction to obtain hidden variables Z (n x a);
step 1.2: determining the number of Gaussian components of the Gaussian mixture model as k according to the number of the working mode varieties of the process, solving the hidden variable Z by using the Gaussian mixture model to obtain the weight, the mean value and the covariance omega = { (pi) of each Gaussian component 1 ,μ 1 ,Σ 1 ),...,(π k ,μ k ,Σ k )}；
Step 1.3: decoding the hidden variable Z by a decoder formed by a single-layer full-connection layer to obtain a decoding result
Calculating reconstruction error e and SPE statistics of samples, and then sequentially obtaining monitoring statistics upper limit SPE of different Gaussian components by using a kernel density estimation method according to clustering results of the samples UCL ；
Step 2: on-line monitoring;
step 2.1: with new sample x of blast furnace new When arriving, the data is normalized, and the normalized data is recorded as x test Then inputting the obtained information into a self-encoder with trained parameters in offline modeling to obtain a reconstruction result
And clustering results of the Gaussian mixture model, and calculating reconstruction errorsDifference e test And calculating the monitoring statistic upper limit SPE according to the probability of the clustering result UCLtest ：
SPE test ＝e test T e test
Wherein the method comprises the steps of
Probability of belonging to the jth gaussian component for the new sample, while SPE UCLj The upper limit of the monitoring statistics for the jth gaussian component;
comparison statistics SPE test And statistics upper limit SPE UCLtest If it is SPE test ≥SPE UCLtest The fault is represented, otherwise, the system is in a normal running state;
if the fault is judged to occur at the moment t, calculating the reconstruction error increment |e of the moment t t -e t-1 And judging the magnitude of the correlation between each variable and the fault according to the magnitude.
2. The method of claim 1, wherein the model uses a gaussian mixture model to spatially cluster data in hidden variables as:
inputting a process variable X (n multiplied by m) into an encoder, namely a GRU model, encoding to obtain a hidden variable Z (n multiplied by a) with a dimension of a (a < m), performing soft clustering in a hidden variable space by using a Gaussian mixture model, setting the clustering number k according to the number of working mode types of the process, and solving through an EM algorithm to obtain weights, mean values and variances of different Gaussian components: Ω= { (pi) 1 ,μ 1 ,Σ 1 ),...,(π k ,μ k ,Σ k ) Of which (pi) i ,μ i ,Σ i ) The weights, means and variances of the ith gaussian component are represented in turn.
3. The method of claim 1, wherein the decoder is a single full link layer.
4. The method of claim 1, wherein the probability-based SPE statistic is:
after the process variable X (n X m) has been reconstructed from the encoder, the output is reconstructed from the input variable X
Calculating reconstruction error e and SPE statistics:
SPE＝e T e
in the offline training stage, the Gaussian mixture model clusters different samples into k Gaussian components, and then calculates the upper limit SPE of the monitoring statistics of the SPE statistics of the samples in each Gaussian component by using a kernel density estimation method UCL The method comprises the steps of carrying out a first treatment on the surface of the The on-line monitoring stage is performed on a new sample x new After arrival, the probability of belonging to different Gaussian components is calculated
Then dynamically calculating monitoring statistic upper limit SPE by probability UCLnew ：
Wherein the method comprises the steps of
Probability of belonging to the jth gaussian component for a new sample.
5. The method according to claim 1, wherein the calculating the correlation degree of different variables and faults according to the reconstruction error increment when the faults occur is as follows: assuming that the moment when the fault occurs is t moment, the reconstruction error of the sample is marked as e t Reconstruction is performedError increment of |e t -e t-1 And judging the correlation between the error and the fault according to the reconstruction error increment of each variable at the moment of occurrence of the fault.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210486776.3A CN114854920B (en) | 2022-05-06 | 2022-05-06 | GRU self-encoder blast furnace abnormity monitoring method embedded by Gaussian mixture model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210486776.3A CN114854920B (en) | 2022-05-06 | 2022-05-06 | GRU self-encoder blast furnace abnormity monitoring method embedded by Gaussian mixture model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114854920A CN114854920A (en) | 2022-08-05 |
| CN114854920B true CN114854920B (en) | 2023-06-30 |
Family
ID=82634747
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210486776.3A Active CN114854920B (en) | 2022-05-06 | 2022-05-06 | GRU self-encoder blast furnace abnormity monitoring method embedded by Gaussian mixture model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114854920B (en) |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106054189B (en) * | 2016-07-17 | 2018-06-05 | 西安电子科技大学 | Radar target identification method based on dpKMMDP models |
| CN106656637B (en) * | 2017-02-24 | 2019-11-26 | 国网河南省电力公司电力科学研究院 | A kind of power grid method for detecting abnormality and device |
| US10604814B2 (en) * | 2017-09-27 | 2020-03-31 | International Business Machines Coporation | Manufacturing process control with deep learning-based predictive model for hot metal temperature of blast furnace |
| CN109558906B (en) * | 2018-11-23 | 2022-05-03 | 四川大学 | Functional magnetic resonance clustering imaging method based on Gaussian hidden variable dimension reduction clustering center |
| CN112488148A (en) * | 2020-10-27 | 2021-03-12 | 清华大学 | Clustering method and device based on variational self-encoder |
| CN114021469A (en) * | 2021-11-15 | 2022-02-08 | 浙江大学 | Method for monitoring one-stage furnace process based on mixed sequence network |
-
2022
- 2022-05-06 CN CN202210486776.3A patent/CN114854920B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN114854920A (en) | 2022-08-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101899563B (en) | PCA (Principle Component Analysis) model based furnace temperature and tension monitoring and fault tracing method of continuous annealing unit | |
| CN112001527A (en) | Industrial production process target data prediction method of multi-feature fusion deep neural network | |
| CN111652414B (en) | Sliding window PCA blast furnace abnormity monitoring method based on Gaussian mixture model | |
| CN104731083B (en) | A kind of industrial method for diagnosing faults and application based on self-adaptive feature extraction | |
| CN110793058B (en) | Method for predicting high-temperature corrosion of boiler water-cooled wall and evaluating service life of boiler water-cooled wall | |
| CN112598144A (en) | CNN-LSTM burst fault early warning method based on correlation analysis | |
| CN102925602A (en) | Furnace profile maintenance method for blast furnace operation | |
| CN109521735B (en) | On-line assessment method for use state risk of high-temperature heating surface of boiler | |
| CN114854920B (en) | GRU self-encoder blast furnace abnormity monitoring method embedded by Gaussian mixture model | |
| CN114997539A (en) | Gas pressure regulator fault prediction method and system | |
| CN112418306B (en) | Gas turbine compressor fault early warning method based on LSTM-SVM | |
| CN113177353A (en) | Data model construction method applied to industrial early warning system | |
| CN108709426B (en) | Sintering machine air leakage fault online diagnosis method based on frequency spectrum characteristic bilateral detection method | |
| CN114004059B (en) | Health portrait method for hydroelectric generating set | |
| CN102494327B (en) | Method for generation retarding and shedding controlling of intra-tube oxide scale of power station boiler tube system | |
| CN103760889B (en) | Fault based on Bayesian network separates fast method | |
| CN106570237B (en) | Turbine stator blade thickness monitoring method and system of blast furnace gas waste heat recovery device | |
| CN114936520A (en) | Steam turbine fault early warning analysis method based on improved MSET | |
| CN111210147B (en) | Sintering process operation performance evaluation method and system based on time sequence feature extraction | |
| CN115146703A (en) | Multi-model-based on-line monitoring method for wall temperature of furnace chamber under wide load of coal-fired unit | |
| CN100371938C (en) | Quality design method under minute new aluminium sample | |
| CN113219942A (en) | Blast furnace fault diagnosis method based on weighted joint distribution adaptive neural network | |
| CN117216697A (en) | Blast furnace ironmaking robust multivariable fault detection method based on time sequence convolution network | |
| Sun et al. | Anomaly detection of marine diesel engine valve system based on Mahalanobis distance | |
| Chen et al. | EPBS_FIDMV: A fault injection and diagnosis methods validation benchmark for EPBS of EMU |
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 |