CN115438897A - Industrial process product quality prediction method based on BLSTM neural network - Google Patents

Industrial process product quality prediction method based on BLSTM neural network Download PDF

Info

Publication number
CN115438897A
CN115438897A CN202210537012.2A CN202210537012A CN115438897A CN 115438897 A CN115438897 A CN 115438897A CN 202210537012 A CN202210537012 A CN 202210537012A CN 115438897 A CN115438897 A CN 115438897A
Authority
CN
China
Prior art keywords
data
training
product quality
variable
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.)
Pending
Application number
CN202210537012.2A
Other languages
Chinese (zh)
Inventor
郭小萍
钟道金
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenyang University of Chemical Technology
Original Assignee
Shenyang University of Chemical Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shenyang University of Chemical Technology filed Critical Shenyang University of Chemical Technology
Priority to CN202210537012.2A priority Critical patent/CN115438897A/en
Publication of CN115438897A publication Critical patent/CN115438897A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06395—Quality analysis or management
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/04—Manufacturing

Abstract

A product quality prediction method of industrial process based on BLSTM neural network relates to an industrial process prediction method, the method of the invention includes the following steps: the method comprises the following steps: analyzing the variable relation of the whole industrial process, and carrying out normalization processing on normal historical data; step two: obtaining reconstructed data by screening the maximum mutual information characteristics of the processed data; step three: substituting the training data divided by the reconstruction data into a bidirectional long-short term memory network to learn the potential relation of the bidirectional long-short term memory network, and optimizing network parameters according to evaluation indexes; step four: when the online quality prediction method is used online, the product quality variable prediction value is obtained through the quality prediction model, the abnormity occurring in the production process is effectively adjusted, and the normal operation of industrial production is ensured. Experiments prove that the method can effectively predict the product quality and has good monitoring effect on the operation state of the industrial process.

Description

Industrial process product quality prediction method based on BLSTM neural network
Technical Field
The invention relates to an industrial process prediction method, in particular to an industrial process product quality prediction method based on a BLSTM neural network.
Background
With the rapid development of national economy and the continuous enlargement of industrial production scale, modern industrial processes develop towards the complicated directions of nonlinearity, unsteady state, high noise, high delay and the like, the product quality is more and more not guaranteed, the loss of manpower and material resources is extremely large, the safety of the industrial processes is also directly related to the national economic development and the safety of lives and properties of people, and the process safety, the product quality, the energy conservation, the emission reduction and the efficiency improvement gradually become the core targets of the modern industry. While new methods and new theories for online analysis and monitoring of key variables of complex industrial processes are continually being proposed, these techniques are overly dependent on accurate model identification and reliable measurement, particularly for online analysis and process monitoring of key process variables. In recent years, quality prediction is widely applied to online evaluation of key product quality with the advantages of high response speed, low maintenance cost, accurate prediction result and the like.
The quality prediction based on data analysis is a common prediction method, such as partial least squares and support vector machines, but the common machine learning network is easily affected by the problem of generalization capability and the problem of gradient disappearance and explosion. Deep neural networks are introduced into predictive modeling by virtue of their better performance, such as deep belief networks, stacked autoencoders, and recurrent neural networks. In order to capture the time dynamic behavior in the time series data, a dynamic recurrent neural network is generated, but the problems of gradient disappearance and gradient explosion still exist. The long-short term memory network is born, the long-short term memory network not only can forget useless information in the past, but also can judge current information and store useful information in a storage unit, but the method cannot solve the problems of different variables on different time step lengths and overlong sequence length, and the quality variable has long time lag characteristics in the industrial process. The bidirectional long and short term memory network is invented to solve the problem, the input sequence is input into the network in a forward and reverse order mode for training, the dependency relationship of the data in the forward and reverse order is mined, the characteristic learning of the bidirectional time sequence is carried out, the characteristic correlation of the time sequence is obtained, and the key characteristic of the time sequence is fully mined.
The noun explains:
the MIC method comprises the following steps: and (4) carrying out mutual information calculation on the maximum mutual information coefficient, namely each related process variable and the product quality variable respectively, and determining the association degree of the process variable and the quality variable through the mutual information number.
BLSTM network: a bidirectional long-short term memory network, a prediction method based on time series.
Disclosure of Invention
The invention aims to provide an industrial process product quality prediction method based on a BLSTM neural network, which aims at extracted complex industrial data, removes data noise and redundancy, strengthens the robustness of a model, deeply excavates the potential relation between relevant process variables of an industrial process time sequence and product quality variables, realizes the real-time prediction of the product quality, timely judges whether the industrial process normally operates, timely avoids operation state faults, improves the production efficiency, reduces the waste of resources, and accurately and visually reflects the operation state of the industrial process.
The purpose of the invention is realized by the following technical scheme:
a product quality prediction method for industrial process based on BLSTM neural network comprises establishing product quality prediction system of training part and testing part;
a training part:
1) Carrying out noise reduction processing on industrial process data, simultaneously dividing a training set and a test set, and then carrying out normalization processing on the training data;
2) The normalized training data is substituted into a maximum mutual information coefficient algorithm, a process variable with a larger correlation coefficient is obtained according to formulas (1) to (4), new recombined training data is formed, and the redundancy of the data is reduced while the screening of the quality-related characteristics is completed;
3) Constructing a bidirectional long-short term memory network model, setting model parameters and the time window length of an input sample, and inputting recombined training data into a network to train the network parameters;
4) Performing RMSE and R2 value evaluation on a prediction result after each training, obtaining optimal parameters of the model by adopting a grid search mode, and obtaining a quality prediction model when an error meets a threshold value;
test part:
1) Extracting the maximum value and the minimum value of each variable of the training data to carry out normalization processing on the test data;
2) The processed test data and the training part complete quality related characteristic screening to obtain recombined test data;
3) And substituting the recombined test data into the quality prediction model to predict the quality variable, and obtaining a product quality variable prediction value through the quality prediction model, thereby effectively adjusting the abnormity appearing in the production process and ensuring the normal operation of industrial production.
The industrial process product quality prediction method based on the BLSTM neural network comprises the following specific steps:
s1: acquiring process data based on an industrial field sensor or acquiring a normal sample X and a product sample Y by an industrial system simulation platform;
s2: deeply knowing the industrial process flow, grasping the relation between relevant process variables and product quality variables, and carrying out noise reduction and normalization processing on the sample;
s3: the related process variable and product quality variable data are subjected to maximum mutual information characteristic screening, low-dimensional reconstruction data of the model are determined, and data redundancy and calculated amount of a neural network are reduced;
s4: and (4) reconstructing data according to the following steps: 1, dividing the data into a training data set and a test data set in proportion;
s5: substituting the reconstructed training data into the bidirectional long-short term memory network to learn the potential relationship of the network, and learning the potential relationship through Root Mean Square Error (RMSE) and R 2 Optimizing parameters of the model by indexes such as coefficients and loss functions (loss) until a quality prediction model is determined;
s6: and substituting the reconstructed test data into the quality prediction model to predict the product, and obtaining a product quality variable prediction value through the quality prediction model, thereby effectively adjusting the abnormity in the production process and ensuring the normal operation of industrial production.
The invention has the advantages and effects that:
1. the stability and robustness of the prediction of the industrial process running state are greatly improved, the bidirectional long-short term memory network can process long-batch data sets and has stronger memory capacity on information, and the prediction can be accurately performed by using the past information and the future information in a good relation. Compared with other methods, the bidirectional long-short term network has stronger generalization capability and robustness on the industrial process state prediction precision, and meets the timeliness requirement of industrial process monitoring prediction.
2. The industrial process data prediction method is more intuitive and reliable, reduces the supervision difficulty of monitoring personnel on the operation of the system, and improves the monitoring efficiency. The invention predicts the quality of the product produced in the operation of the industrial process, obtains the predicted value of the quality variable of the product through the quality prediction model when in online operation, effectively adjusts the abnormity appearing in the production process and ensures the normal operation of the industrial production.
Drawings
FIG. 1 is a plot of the loss function of various models of the present invention;
FIG. 2 is a comparison of test samples of the present invention;
FIG. 3 is a graph of predicted results for different model quality variables according to the present invention;
FIG. 4 is an overall flow diagram of the present invention
FIG. 5 is a diagram of a bidirectional long short term memory network according to the present invention.
Detailed Description
The present invention will be described in detail with reference to the embodiments shown in the drawings.
The invention relates to an industrial process product quality prediction method based on a BLSTM neural network, which is used for knowing the position setting of a sensor in the whole industrial process and how acquired data is utilized. The method utilizes the maximum mutual information coefficient to screen the characteristic variables, removes the redundancy of data and reduces the network calculation amount. Meanwhile, the method carries out grid search mode optimization on the parameters of the bidirectional long-short term memory network, and establishes a complete product quality prediction system, and the whole steps of the prediction method are divided into a training part and a testing part.
The method comprises the following steps:
s1: acquiring process data based on an industrial field sensor or acquiring a normal sample X and a product sample Y by an industrial system simulation platform;
s2: deeply knowing the industrial process flow, grasping the relation between relevant process variables and product quality variables, and carrying out noise reduction and normalization processing on the sample;
s3: the related process variable and product quality variable data are subjected to maximum mutual information characteristic screening, low-dimensional reconstruction data of the model are determined, and data redundancy and calculated amount of a neural network are reduced;
s4: and (4) reconstructing data according to the following steps: 1, dividing the data into a training data set and a test data set in proportion;
s5: substituting the reconstructed training data into the bidirectional long-short term memory network to learn the potential relationship of the network, and learning the potential relationship through Root Mean Square Error (RMSE) and R 2 Optimizing parameters of the model by indexes such as coefficients and loss functions (loss) until a quality prediction model is determined;
s6: and substituting the reconstructed test data into the quality prediction model to predict the product, and obtaining a product quality variable prediction value through the quality prediction model, thereby effectively adjusting the abnormity in the production process and ensuring the normal operation of industrial production.
Further improvement, the step S3 of screening the maximum mutual information coefficient characteristics of the related process variables and the product quality variables specifically includes the following steps:
s31: the MIC probability equation is as follows:
Figure RE-DEST_PATH_IMAGE002
Figure RE-DEST_PATH_IMAGE004
is a variable of
Figure RE-DEST_PATH_IMAGE006
And
Figure RE-DEST_PATH_IMAGE008
the MIC algorithm is characterized in that the two variables are dispersed in a two-dimensional space according to the relation between the two variables, a scatter diagram is used for representing the relation, the current two-dimensional space is divided into certain interval numbers in the x direction and the y direction respectively, and then the current scatter point in each interval number is checkedAnd under the condition that the grid falls into the grid, joint probability calculation is utilized, so that the problem that joint probability in mutual information is difficult to obtain is solved.
S32: the MIC screening formula is as follows:
Figure RE-DEST_PATH_IMAGE010
in the above formula
Figure RE-DEST_PATH_IMAGE012
，
Figure RE-DEST_PATH_IMAGE014
The number of the grids divided in the x and y directions is essentially the grid distribution,
Figure RE-DEST_PATH_IMAGE016
is a variable.
In a further improvement, the step S4 of dividing the industrial process data into training data and test data includes the following specific steps:
s4: the method is verified by applying a Tennessee Eastman industrial process data case, and the acquired process data consists of three parts, namely a process control variable, a process measurement variable and a component measurement variable, wherein 47 process variables are acquired to predict 5 product variables which are difficult to measure, experimental data is formed by combining 5 times of TE normal industrial process data, corresponding reconstruction is performed according to nodes at the same time, 4800 sample data are reconstructed for each variable, and the variables are divided into a training set part and a test set part, wherein the training set part comprises the first 4000 samples, and the test set part comprises the last 800 samples; secondly, by knowing the whole process, the external relation among the variables is expected, and the deep relation is further mined.
Further improvement, the bidirectional long-short term memory network used by the invention solves the problem that the neural network is difficult to model a long sequence, and can also keep the robustness and stability of the model performance; the context information is perfectly combined by utilizing two forward and reverse long-term and short-term memory networks in the network, and the characteristic feature mining of deep key variables is realized, so that more accurate prediction is achieved. The step S5 of using the bidirectional long and short term memory network parameter tuning process to obtain the predicted value Y includes the following specific steps:
s51, substituting training data into the bidirectional long and short term memory network training according to the divided training data and test data;
s52: the structure of the bidirectional long-short term memory network is shown in fig. 5: a forward layer: the formula for updating the neural network layer from left to right is
Figure RE-DEST_PATH_IMAGE018
Reverse layer: the formula for updating the neural network layer from right to left is
Figure RE-DEST_PATH_IMAGE020
An output layer: the two layers of the circulating neural network layers are superposed and then output as
Figure RE-DEST_PATH_IMAGE022
In the formula:
Figure RE-DEST_PATH_IMAGE024
is composed of
Figure RE-DEST_PATH_IMAGE026
Hidden layer vectors in the forward direction of time;
Figure RE-926251DEST_PATH_IMAGE026
is a time node;
Figure RE-DEST_PATH_IMAGE028
is a time of day
Figure RE-30342DEST_PATH_IMAGE026
The input of (1);
Figure RE-DEST_PATH_IMAGE030
is a time of day
Figure RE-747762DEST_PATH_IMAGE026
An output of time;
Figure RE-DEST_PATH_IMAGE032
a weight matrix that is an input-hidden layer;
Figure RE-DEST_PATH_IMAGE034
a weight matrix of hidden layer-hidden layer;
Figure RE-DEST_PATH_IMAGE036
a weight matrix of hidden layer-output layer;
Figure RE-DEST_PATH_IMAGE038
is a hidden layer bias vector;
Figure RE-DEST_PATH_IMAGE040
is the output layer bias vector;
Figure RE-DEST_PATH_IMAGE042
activating a function for the hidden layer; the arrow above the parameter symbol represents the direction; in the network model, an Adam function is used as an activation function, and a Dropout function is used on the network to discard a part of neural network nodes, so that overfitting of the network is prevented;
s53: the training data is substituted into the model to optimize each internal parameter in a grid search mode until the obtained prediction result reaches the best, the loss function reaches the minimum and the evaluation index of the model reaches the qualified quality prediction model;
s54: and substituting the test data into the quality prediction model to predict the result and calculate the evaluation index, thereby realizing the judgment of whether the operation state of the industrial process is normal.
Further improvement, the evaluation indexes used in the step S5 are less, and the model can be evaluated through more evaluation indexes, so that the model can be evaluated more comprehensively, the prediction result can be more accurate, and the industrial process can be monitored substantially better.
The model predicted quality variable process is as follows:
a training part:
1) Carrying out noise reduction processing on industrial process data, simultaneously dividing a training set and a test set, and then carrying out normalization processing on the training data;
2) The normalized training data is substituted into a maximum mutual information coefficient algorithm, a process variable with a larger correlation coefficient is obtained according to formulas (1) to (4), new recombined training data is formed, and the redundancy of the data is reduced while the screening of the quality-related characteristics is completed;
3) Constructing a bidirectional long-short term memory network model, setting model parameters and the time window length of an input sample, and inputting recombined training data into a network to train the network parameters;
4) Performing RMSE and R2 value evaluation on a prediction result after each training, obtaining optimal parameters of the model by adopting a grid search mode, and obtaining a quality prediction model when an error meets a threshold value;
test part:
1) Extracting the maximum value and the minimum value of each variable of the training data to carry out normalization processing on the test data;
2) The processed test data and the training part 2) complete quality related characteristic screening to obtain recombined test data;
3) And substituting the recombined test data into the quality prediction model to predict the quality variable, and obtaining a product quality variable predicted value through the quality prediction model, thereby effectively adjusting the abnormity appearing in the production process and ensuring the normal operation of industrial production.
Example (b):
the Tennessee Eastman process is a simulation platform which is developed by Eastman chemical companies and is simulated according to actual chemical reaction processes. The data collected in the process comprises three major parts, namely process control variables, process measurement variables and component measurement variables, wherein 47 process variables (with variable numbers of XMEAS (1) -XMEAS (36) and XMV (1) -XMV (11)) are adopted to predict 5 product variables (with variable numbers of XMEAS (37) -XMEAS (41)) which are difficult to measure, the experimental data in the process is formed by combining 5 times of TE normal industrial process data, corresponding reconstruction is carried out according to nodes at the same time, 4800 sample data are reconstructed for each variable, and the variable is divided into a training set part and a testing set part, wherein the training set part is the first 4000 samples, and the testing set part is the last 800 samples.
1. The evaluation indexes adopted by the invention are as follows:
root Mean Square Error (RMSE):
Figure RE-DEST_PATH_IMAGE044
in the formula:
Figure RE-DEST_PATH_IMAGE046
is the number of the samples, and the number of the samples,
Figure RE-DEST_PATH_IMAGE048
is the true value of the,
Figure RE-DEST_PATH_IMAGE050
is a predicted value, the smaller the RMSE, the better.
R 2 Coefficient:
Figure RE-DEST_PATH_IMAGE052
in the formula:
Figure RE-516873DEST_PATH_IMAGE046
is the number of the samples, and the number of the samples,
Figure RE-518196DEST_PATH_IMAGE048
is the true value of the,
Figure RE-340658DEST_PATH_IMAGE050
is a predicted value of the number of the frames,
Figure RE-DEST_PATH_IMAGE054
is the average value, R 2 The results are normalized, and the difference between models can be more easily seen.
TABLE 1 prediction of product quality by different models
Figure RE-DEST_PATH_IMAGE056
As can be seen from the table 1 and the drawings in the specification, the prediction results of the traditional long-short term memory network and the two-way long-short term memory network on the data such as quality and quantity in the industrial process are not very accurate, and the root mean square error and R are not very accurate 2 The prediction error is larger in the evaluation indexes such as the coefficient, and the long-short term memory network can only simply predict a short batch, so that the method has great limitation; the bidirectional long-short term memory network can link context information to mine deep potential information, and can also perform prediction on large batches of data well, but the model prediction is seriously interfered by too many inconsequential variables of industrial process data, so that the prediction effect is poor; the invention utilizes the maximum mutual information coefficient to screen the characteristic variables of the model for dimension reduction and reconstruction of input data, so that the robustness and generalization capability of the model are enhanced.
While embodiments of the invention have been disclosed above, it is not limited to the applications set forth in the specification and the embodiments, which are fully applicable to various fields of endeavor for which the invention pertains, and further modifications may readily be made by those skilled in the art, it being understood that the invention is not limited to the details shown and described herein without departing from the general concept defined by the appended claims and their equivalents.

Claims (2)

1. The industrial process product quality prediction method based on the BLSTM neural network is characterized by comprising the steps of establishing a product quality prediction system of a training part and a testing part;
a training part:
1) Carrying out noise reduction processing on industrial process data, simultaneously dividing a training set and a test set, and then carrying out normalization processing on the training data;
2) The normalized training data is substituted into a maximum mutual information coefficient algorithm, a process variable with a larger correlation coefficient is obtained according to formulas (1) to (4), new recombined training data is formed, and the redundancy of the data is reduced while the screening of the quality-related characteristics is completed;
3) Constructing a bidirectional long-short term memory network model, setting model parameters and the time window length of an input sample, and inputting recombined training data into a network to train the network parameters;
4) Performing RMSE and R2 value evaluation on a prediction result after each training, obtaining optimal parameters of the model by adopting a grid searching mode, and obtaining a quality prediction model when an error meets a threshold value;
and a test part:
1) Extracting the maximum value and the minimum value of each variable of the training data to carry out normalization processing on the test data;
2) The processed test data and the training part complete quality-related characteristic screening to obtain recombined test data;
3) And substituting the recombined test data into the quality prediction model to predict the quality variable, and obtaining a product quality variable predicted value through the quality prediction model, thereby effectively adjusting the abnormity appearing in the production process and ensuring the normal operation of industrial production.
2. The industrial process product quality prediction method based on the BLSTM neural network as claimed in claim 1, wherein the method comprises the following steps:
s1, acquiring process data based on an industrial field sensor or acquiring a normal sample X and a product sample Y by an industrial system simulation platform;
s2, deeply knowing the industrial process flow, grasping the relation between relevant process variables and product quality variables, and carrying out noise reduction and normalization processing on the sample;
s3, screening the maximum mutual information characteristics of the related process variable data and the product quality variable data, determining the low-dimensional reconstruction data of the model, and reducing the redundancy of the data and the calculated amount of a neural network;
and S4, reconstructing data according to the ratio of 4:1, dividing the data into a training data set and a test data set in proportion;
s5, replacing the reconstructed training dataLearning the potential relationship in the bidirectional long-short term memory network by Root Mean Square Error (RMSE) and R 2 Optimizing parameters of the model by indexes such as coefficients and loss functions (loss) until a quality prediction model is determined;
and S6, substituting the reconstructed test data into the quality prediction model to predict the product, obtaining a product quality variable prediction value through the quality prediction model, effectively adjusting the abnormity in the production process and ensuring the normal operation of industrial production.
CN202210537012.2A 2022-05-18 2022-05-18 Industrial process product quality prediction method based on BLSTM neural network Pending CN115438897A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210537012.2A CN115438897A (en) 2022-05-18 2022-05-18 Industrial process product quality prediction method based on BLSTM neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210537012.2A CN115438897A (en) 2022-05-18 2022-05-18 Industrial process product quality prediction method based on BLSTM neural network

Publications (1)

Publication Number Publication Date
CN115438897A true CN115438897A (en) 2022-12-06

Family

ID=84240899

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210537012.2A Pending CN115438897A (en) 2022-05-18 2022-05-18 Industrial process product quality prediction method based on BLSTM neural network

Country Status (1)

Country Link
CN (1) CN115438897A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116052789A (en) * 2023-03-29 2023-05-02 河北大景大搪化工设备有限公司 Toluene chlorination parameter automatic optimization system based on deep learning
CN116312861A (en) * 2023-05-09 2023-06-23 济南作为科技有限公司 Denitration system gas concentration prediction method, device, equipment and storage medium

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116052789A (en) * 2023-03-29 2023-05-02 河北大景大搪化工设备有限公司 Toluene chlorination parameter automatic optimization system based on deep learning
CN116052789B (en) * 2023-03-29 2023-09-15 河北大景大搪化工设备有限公司 Toluene chlorination parameter automatic optimization system based on deep learning
CN116312861A (en) * 2023-05-09 2023-06-23 济南作为科技有限公司 Denitration system gas concentration prediction method, device, equipment and storage medium

Similar Documents

Publication Publication Date Title
CN111222549B (en) Unmanned aerial vehicle fault prediction method based on deep neural network
CN115438897A (en) Industrial process product quality prediction method based on BLSTM neural network
CN113298297B (en) Wind power output power prediction method based on isolated forest and WGAN network
CN112181706B (en) Power dispatching data anomaly detection method based on logarithmic interval isolation
CN113065703A (en) Time series prediction method combining multiple models
Sun et al. Data-driven modeling and analysis based on complex network for multimode recognition of industrial processes
CN112949201B (en) Wind speed prediction method and device, electronic equipment and storage medium
CN112101482B (en) Method for detecting abnormal parameter mode of missing satellite data
CN113807951A (en) Transaction data trend prediction method and system based on deep learning
CN112731890A (en) Power plant equipment fault detection method and device
CN114943328A (en) SARIMA-GRU time sequence prediction model based on BP neural network nonlinear combination
CN115496264A (en) Method for predicting generated power of wind turbine generator
CN115935814A (en) Transformer fault prediction method based on ARIMA-SVM model
CN112184037B (en) Multi-modal process fault detection method based on weighted SVDD
CN114818493A (en) Method for quantitatively evaluating integrity degree of tunnel rock mass
CN114548494A (en) Visual cost data prediction intelligent analysis system
CN114676887A (en) River water quality prediction method based on graph convolution STG-LSTM
CN113988210A (en) Method and device for restoring distorted data of structure monitoring sensor network and storage medium
Song et al. Sparse adjacency forecasting and its application to efficient root cause diagnosis of process faults
Yan et al. Nonlinear quality-relevant process monitoring based on maximizing correlation neural network
Zheng et al. Research on Predicting Remaining Useful Life of Equipment Based on Health Index
CN117647725B (en) Aging test method and system for PCBA
Liu et al. Prediction of hydraulic pumps remaining useful life based on LSTM and Transform with dual self-attention
Xiaolong et al. Research on Fault Diagnosis Method of Rolling Bearing Based on Improved Convolutional Neural Network
CN117829822A (en) Power transformer fault early warning method and system

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