CN112612782A - MES system data online filling method and system based on LSTM network - Google Patents

MES system data online filling method and system based on LSTM network Download PDF

Info

Publication number
CN112612782A
CN112612782A CN202011500119.7A CN202011500119A CN112612782A CN 112612782 A CN112612782 A CN 112612782A CN 202011500119 A CN202011500119 A CN 202011500119A CN 112612782 A CN112612782 A CN 112612782A
Authority
CN
China
Prior art keywords
data
lstm network
prediction model
data prediction
network data
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
CN202011500119.7A
Other languages
Chinese (zh)
Inventor
柴森春
王昭洋
张佳
张百海
崔灵果
李慧芳
姚分喜
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.)
Beijing Institute of Technology BIT
Original Assignee
Beijing Institute of Technology BIT
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 Beijing Institute of Technology BIT filed Critical Beijing Institute of Technology BIT
Priority to CN202011500119.7A priority Critical patent/CN112612782A/en
Publication of CN112612782A publication Critical patent/CN112612782A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/21—Design, administration or maintenance of databases
    • G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield 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
    • G06N3/045—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
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biophysics (AREA)
  • Evolutionary Computation (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Databases & Information Systems (AREA)
  • Quality & Reliability (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses an MES system data online filling method and system based on an LSTM network, wherein the method comprises the following steps: obtaining an LSTM network data prediction model, wherein the LSTM network data prediction model is a trained LSTM network model; when the field data is not received at the current time T, the time sequence data [ X ] is transmittedT‑1,XT‑2,…,XT‑j,…,XT‑N]Inputting the LSTM network data prediction model, and predicting the data information at the current time T online, wherein j is 1,2, …, N, XT‑jFor the data information X at the current momentTData information of previous j time steps; when the field data is received in the current time interval T, the time sequence data [ X ] is usedT‑1,XT‑2,…,XT‑j,…,XT‑N]Received as input at the current time TAnd the field data is a label, and the LSTM network data prediction model is trained and updated on line. The invention realizes the on-line filling of the data and simultaneously ensures the accuracy of the data filling.

Description

MES system data online filling method and system based on LSTM network
Technical Field
The invention relates to the field of steel processing and manufacturing, in particular to an MES system data online filling method and system based on an LSTM network.
Background
In the field of steel production, it is necessary to monitor parameters in the production process, such as temperature, flow, pressure, voltage, current, etc., in order to monitor the production process, the operating state of the machine, etc. The data are generally obtained by a sensor and are time-series data transmitted according to time, meanwhile, in the acquisition stage, due to objective reasons, an information missing phenomenon often occurs, and the missing of the data can have adverse effects on the operation of subsequent data and the storage of a time-series database, so that the missing data is very necessary to be filled.
Most of the existing data missing filling methods are off-line filling methods, namely, after the data are stored in a database, the missing data are judged and filled. According to the off-line filling method, when the data volume in the database is large, the delay of data utilization is large, and the problem of low reliability caused by the fact that the data cannot be directly utilized after being acquired (the data needs to be transmitted to the database for filling) also exists. For some special industrial sites, when data loss occurs, subsequent machines may not be able to maintain normal operation, which makes the off-line data filling method unable to meet the above requirements.
Disclosure of Invention
The invention aims to provide an MES system data online filling method and system based on an LSTM network.
In order to achieve the purpose, the invention provides the following scheme:
an MES system data online filling method based on an LSTM network comprises the following steps:
step 1: obtaining an LSTM network data prediction model, wherein the LSTM network data prediction model is a trained LSTM network model;
step 2: when the field data is not received at the current time T, the time sequence data [ X ] is transmittedT-1,XT-2,…,XT-j,…,XT-N]Inputting the LSTM network data prediction model, and predicting the data information at the current time T online, wherein j is 1,2, …, N, XT-jFor the data information X at the current momentTData information of previous j time steps;
and step 3: when the field data is received in the current time interval T, the time sequence data [ X ] is usedT-1,XT-2,…,XT-j,…,XT-N]And taking the field data received at the current time T as a label for input, and carrying out online training and updating on the LSTM network data prediction model.
Optionally, the online training of the LSTM network data prediction model specifically includes:
normalizing the time sequence data and the field data received at the current time T;
and training the LSTM network data prediction model by taking the normalized time sequence data as input and the normalized field data as a label.
Optionally, the online prediction of the data information of the current time T specifically includes:
normalizing the time series data;
inputting the normalized time sequence data into the LSTM network data prediction model;
and performing inverse normalization on the output result of the LSTM network data prediction model to obtain data information of the current time T.
Optionally, the data information includes a plurality of dimensions, and the normalization of the data information is normalization of each dimension.
Optionally, when the LSTM network data prediction model is not updated, the LSTM network data prediction model obtained in step 1 is a pre-trained LSTM model; and (3) when the LSTM network data prediction model is updated, the LSTM network data prediction model obtained in the step (1) is the updated LSTM network data prediction model.
Optionally, in step 3, a learning rate used when the LSTM network data prediction model is trained is lower than a learning rate used when the LSTM network data prediction model is pre-trained.
The invention also provides an MES system data online filling system based on the LSTM network, which comprises the following steps:
the data prediction model acquisition module is used for acquiring an LSTM network data prediction model, and the LSTM network data prediction model is a trained LSTM network model;
a filling data prediction module for predicting the time sequence data [ X ] when the field data is not received at the current time TT-1,XT-2,…,XT-j,…,XT-N]Inputting the LSTM network data prediction model, and predicting the data information at the current time T online, wherein j is 1,2, …, N, XT-jFor the data information X at the current momentTData information of previous j time steps;
a data prediction model updating module for updating the time sequence data [ X ] when the field data is received in the current time period TT-1,XT-2,…,XT-j,…,XT-N]And taking the field data received at the current time T as a label for input, and carrying out online training and updating on the LSTM network data prediction model.
Alternatively to this, the first and second parts may,
the filling data prediction module specifically comprises:
the normalization unit is used for normalizing the time sequence data;
the filling data prediction unit is used for inputting the normalized time sequence data into the LSTM network data prediction model;
the inverse normalization unit is used for performing inverse normalization on the output result of the LSTM network data prediction model to obtain data information of the current time T;
the data prediction model updating module specifically comprises:
the normalization unit is used for normalizing the time sequence data and the field data received at the current time T;
and the data prediction model training unit is used for training the LSTM network data prediction model by taking the normalized time sequence data as input and the normalized field data as a label.
Optionally, the data information includes a plurality of dimensions, and the normalization of the data information is normalization of each dimension.
Optionally, when the LSTM network data prediction model is not updated, the LSTM network data prediction model acquired by the data prediction model acquisition module is a pre-trained LSTM model; when the LSTM network data prediction model is updated, the LSTM network data prediction model acquired by the data prediction model acquisition module is the updated LSTM network data prediction model; and the learning rate adopted by the data prediction model updating module when the LSTM network data prediction model is trained is lower than the learning rate adopted by the LSTM network data prediction model when the LSTM network data prediction model is pre-trained.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects: according to the method and the system for online filling of the MES system data based on the LSTM network, when the MES system does not collect the data collected by the sensor at the current moment, the data is online predicted and filled by adopting the LSTM network data prediction model.
In addition, the invention adopts the LSTM network data prediction model to carry out prediction filling on the data, thereby improving the accuracy of filling the missing data. The invention also carries out real-time online learning on the LSTM network data prediction model, when the actual data changes, the LSTM network data prediction model can be correspondingly adjusted, thereby improving the accuracy of the LSTM network data prediction model and further improving the accuracy of missing data filling.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
Fig. 1 is a schematic flow chart of an online filling method for data of an MES system based on an LSTM network according to embodiment 1 of the present invention;
fig. 2 is a schematic structural diagram of an online MES system data shimming system based on an LSTM network according to embodiment 2 of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Example 1
Referring to fig. 1, the embodiment provides an online filling method for MES system data based on an LSTM network, which includes the following steps:
step 1: and obtaining an LSTM network data prediction model, wherein the LSTM network data prediction model is a trained LSTM network model. In this embodiment, the LSTM network is pre-trained in advance by using some existing time series data in the steel processing process, for example, the temperature time series data of a certain link in the steel processing and manufacturing process, and the LSTM network is pre-trained by using these historical time series data to obtain an LSTM network data prediction model for filling up the temperature data of the link in the following process.
Step 2: when the field data is not received at the current time T, the time sequence data [ X ] is transmittedT-1,XT-2,…,XT-j,…,XT-N]Inputting the LSTM network data prediction model, and predicting the data information at the current time T online, wherein j is 1,2, …, N, XT-jFor the data information X at the current momentTData information of previous j time steps.
And step 3: when the field data is received in the current time interval T, the time sequence data [ X ] is usedT-1,XT-2,…,XT-j,…,XT-N]And taking the field data received at the current time T as a label for input, and carrying out online training and updating on the LSTM network data prediction model.
In the present embodiment, the field data is temperature sequence information of the iron-making blast furnace in the steel manufacturing process, and in other embodiments, the field data may be temperature sequence information of the steel-making converter in the steel manufacturing process, temperature sequence information of the refining furnace, pressure sequence information of the withdrawal and straightening unit in the continuous casting process, or speed sequence information of the rotary table.
When on-line data filling is carried out, the LSTM network data prediction model called for the first time is the pre-trained LSTM network model. After the LSTM network data prediction model is obtained, whether industrial field data, namely transmission of sensing data exists in the current time period T is judged, wherein T is a sampling period of an industrial field sensor. If the sensing data are transmitted in the time period T, the data are not lost in the current time period, and the data do not need to be processed; at the moment, the currently transmitted data is taken as a training sample, the LSTM network data prediction model is trained and updated, the fine-tuned network model parameters are stored, and the LSTM network data prediction model which is updated latest is called when the LSTM network data prediction model is called next time. If no sensing data is transmitted in the time period T, the data loss occurs in the current time period, the data at the moment needs to be filled, and the current output result of the LSTM network data prediction model is taken as the filling data of the current missing data.
In this embodiment, the structure of the LSTM network data prediction model includes an input layer, a hidden layer, and an output layer, where an input time step of the input layer is N, which represents data input at current N moments selected by a user, the number of layers of the hidden layer and the number of LSTM units of the corresponding layer may select some typical values, the general principle of selection is deep and narrow, and a step of the LSTM unit of the output layer is 1, which represents prediction output at the current moment. The input data of the current N continuous moments are used as the input of the LSTM network data prediction model, the output of the LSTM network data prediction model is the predicted value of the industrial field data at the current moment, the input data format of the network model is (m, N), wherein m is the dimension of industrial field data transmission, and the output data format of the network model is (m, 1).
When the model is used for the first time, the prior industrial field data is required to be utilized to pre-train the LSTM network data prediction model so that the LSTM network data prediction model can work normally, and the training method of the model is the same as that of a general neural network training method, namely, a gradient descent algorithm of reverse transmission is adopted for training; due to the unique gate structure of the LSTM unit, the problem of gradient disappearance or gradient explosion of the LSTM network data prediction model in the training process can be avoided; because the LSTM network data prediction model can perform online updating of model parameters in the working process, stored model parameters can be directly loaded for later use without repeated pre-training of the model, namely, a large amount of time is spent for offline pre-training of the model except for the first use of the model, and only the stored online learning model parameter information needs to be loaded during subsequent use.
In step 2, when predicting the data information of the current time T on line, firstly normalizing the time series data; then, inputting the normalized time sequence data into the LSTM network data prediction model; and finally, performing inverse normalization on the output result of the LSTM network data prediction model to obtain data information of the current time T. In step 3, when the LSTM network data prediction model is trained on line, firstly, the time sequence data and the field data received at the current time T are normalized; and then, training an LSTM network data prediction model by taking the normalized time sequence data as input and the normalized field data as a label. The specific process may be as follows:
(1) and (4) normalizing the data. Data XT，XT-1，XT-2，…，XT-N(XT-jData X representing the current timeTData information of previous j time steps), the normalization formula is as follows:
Figure BDA0002843355750000061
wherein the content of the first and second substances,
Figure BDA0002843355750000062
for the data after normalization, XmaxAnd XminAnd the data vector is formed by the maximum value and the minimum value in the corresponding dimensionality of the data. In this embodiment, the data information includes a plurality of dimensions, and the normalization of the data information is normalization of each dimension, that is:
XT-i＝(XT-i，1,XT-i，2,L,XT-i，j)(j＝1,2,L,m)
wherein, XT-i，jData X representing the current timeTThe j-th dimension of the previous i time steps, then XmaxAnd XminCan be expressed as follows:
Xmax＝(X，1max,X，2max,L,X,jmax)(j＝1,2,L,m)
Xmin＝(X，1min,X，2min,L,X,jmin)(j＝1,2,L,m)
wherein. X,jmaxAnd X,jminThe maximum value and the minimum value in the corresponding dimension of the data are obtained.
(2) And (4) forward transmission. Mixing XT-1，XT-2，…，XT-NData after normalization
Figure BDA0002843355750000071
Figure BDA0002843355750000072
As the input of the LSTM network data prediction model, the forward transmission training of the model is carried out to obtain the predicted value of the forward transmission of the model
Figure BDA0002843355750000073
If the prediction is on-line, the final prediction filling value needs to be subjected to inverse normalization processing till the current step, namely the final filling value
Figure BDA0002843355750000074
Comprises the following steps:
Figure BDA0002843355750000075
(3) a loss function is calculated. If the model is updated on line, the loss function is calculated, because only the data at the current moment is used as a training sample during on-line training, namely only one piece of training data is used for carrying out fine adjustment on the model parameter, the loss function ylossThe calculation formula is defined as follows:
Figure BDA0002843355750000076
(4) and (4) reverse transmission. And carrying out reverse transmission of the error by using a gradient descent method.
In this embodiment, when the LSTM network data prediction model is trained online, the learning rate used is lower than that used when the LSTM network data prediction model is pre-trained, so that the phenomenon that the LSTM network data prediction model is over-fitted to the data at the current time is avoided. Since a large amount of data is subsequently brought to the LSTM network data prediction model for online learning, the LSTM network will slowly learn the change even if the actual data model changes at a low learning rate.
Example 2
Referring to fig. 2, the embodiment provides an online MES system data filling system based on an LSTM network, which includes:
a data prediction model obtaining module 201, configured to obtain an LSTM network data prediction model, where the LSTM network data prediction model is a trained LSTM network model;
a padding data prediction module 202 for predicting the time-series data [ X ] when the field data is not received at the current time TT-1,XT-2,…,XT-j,…,XT-N]Inputting the LSTM network data prediction model, and predicting the data information at the current time T online, wherein j is 1,2, …, N, XT-jFor the data information X at the current momentTData information of previous j time steps;
a data prediction model updating module 203 for updating the time series data [ X ] when the field data is received in the current time interval TT-1,XT-2,…,XT-j,…,XT-N]And taking the field data received at the current time T as a label for input, and carrying out online training and updating on the LSTM network data prediction model.
The padding data prediction module 202 specifically includes:
the normalization unit is used for normalizing the time sequence data;
the filling data prediction unit is used for inputting the normalized time sequence data into the LSTM network data prediction model;
the inverse normalization unit is used for performing inverse normalization on the output result of the LSTM network data prediction model to obtain data information of the current time T;
the data prediction model updating module 203 specifically includes:
the normalization unit is used for normalizing the time sequence data and the field data received at the current time T;
and the data prediction model training unit is used for training the LSTM network data prediction model by taking the normalized time sequence data as input and the normalized field data as a label.
The data information comprises a plurality of dimensions, and the normalization of the data information is the normalization of each dimension. When the LSTM network data prediction model is not updated, the LSTM network data prediction model acquired by the data prediction model acquisition module is a pre-trained LSTM model; when the LSTM network data prediction model is updated, the LSTM network data prediction model acquired by the data prediction model acquisition module is the updated LSTM network data prediction model; and the learning rate adopted by the data prediction model updating module when the LSTM network data prediction model is trained is lower than the learning rate adopted by the LSTM network data prediction model when the LSTM network data prediction model is pre-trained.
Compared with the existing data missing filling method, the method has the following advantages:
(1) the data filling precision is higher. On one hand, the LSTM network is used for modeling the data model, has long and short term memory performance, can memorize some previous data characteristics, and improves the accuracy of missing data filling; meanwhile, the LSTM network data prediction model can be used for learning the nonlinear relation in the actual data model, the error of the established model is smaller than that of the traditional model, and the accuracy of data missing and filling is further improved. On the other hand, through online learning of the LSTM network, when the actual data model changes, the LSTM network can capture the change and perform corresponding adjustment, so that the accuracy and the real-time performance of the model are relatively higher, and the accuracy of missing data filling is improved.
(2) The invention belongs to online filling. When data loss occurs, the loss detection and filling are carried out immediately, so that whether the data is uploaded to a time sequence database or subsequent transmission operation is carried out subsequently, the processed data is directly complete and has no loss, the data availability is improved, and the cost of carrying out subsequent processing in the database is reduced.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.

Claims (10)

1. An MES system data online filling method based on an LSTM network is characterized by comprising the following steps:
step 1: obtaining an LSTM network data prediction model, wherein the LSTM network data prediction model is a trained LSTM network model;
step 2: when the field data is not received at the current time T, the time sequence data [ X ] is transmittedT-1,XT-2,…,XT-j,…,XT-N]Inputting the LSTM network data prediction model, and predicting the data information at the current time T online, wherein j is 1,2, …, N, XT-jFor the data information X at the current momentTData information of previous j time steps;
and step 3: when the field data is received in the current time interval T, the time sequence data [ X ] is usedT-1,XT-2,…,XT-j,…,XT-N]And taking the field data received at the current time T as a label for input, and carrying out online training and updating on the LSTM network data prediction model.
2. The LSTM network based MES system data online filling method of claim 1, wherein online training of the LSTM network data prediction model specifically comprises:
normalizing the time sequence data and the field data received at the current time T;
and training the LSTM network data prediction model by taking the normalized time sequence data as input and the normalized field data as a label.
3. The LSTM network-based MES system data online filling method of claim 1, wherein the online predicting of data information at a current time T specifically includes:
normalizing the time series data;
inputting the normalized time sequence data into the LSTM network data prediction model;
and performing inverse normalization on the output result of the LSTM network data prediction model to obtain data information of the current time T.
4. An LSTM network based MES system data online filling method according to claim 2 or 3, wherein said data information comprises several dimensions, and the normalization of the data information is a normalization of each dimension.
5. The LSTM network based MES system data online filling method of claim 1, wherein when the LSTM network data prediction model is not updated, the LSTM network data prediction model obtained in step 1 is a pre-trained LSTM model; and (3) when the LSTM network data prediction model is updated, the LSTM network data prediction model obtained in the step (1) is the updated LSTM network data prediction model.
6. The LSTM network based MES system data online filling method of claim 5, wherein the LSTM network data prediction model is trained in step 3 at a lower learning rate than the LSTM network data prediction model was pre-trained.
7. An MES system data online filling system based on an LSTM network, which is characterized by comprising:
the data prediction model acquisition module is used for acquiring an LSTM network data prediction model, and the LSTM network data prediction model is a trained LSTM network model;
a filling data prediction module for predicting the time sequence data [ X ] when the field data is not received at the current time TT-1,XT-2,…,XT-j,…,XT-N]Inputting the LSTM network data prediction model, and predicting the data information at the current time T online, wherein j is 1,2, …, N, XT-jFor the data information X at the current momentTData information of previous j time steps;
a data prediction model updating module for updating the time sequence data [ X ] when the field data is received in the current time period TT-1,XT-2,…,XT-j,…,XT-N]And taking the field data received at the current time T as a label for input, and carrying out online training and updating on the LSTM network data prediction model.
8. An LSTM network based MES system data in-line shimming system according to claim 7,
the filling data prediction module specifically comprises:
the normalization unit is used for normalizing the time sequence data;
the filling data prediction unit is used for inputting the normalized time sequence data into the LSTM network data prediction model;
the inverse normalization unit is used for performing inverse normalization on the output result of the LSTM network data prediction model to obtain data information of the current time T;
the data prediction model updating module specifically comprises:
the normalization unit is used for normalizing the time sequence data and the field data received at the current time T;
and the data prediction model training unit is used for training the LSTM network data prediction model by taking the normalized time sequence data as input and the normalized field data as a label.
9. The LSTM network-based MES system data in-line shimming system of claim 8, wherein the data information comprises a plurality of dimensions, and the normalization of the data information is a normalization of the dimensions.
10. The LSTM network based MES system data online shimming system of claim 7, wherein the LSTM network data prediction model obtained by the data prediction model obtaining module is a pre-trained LSTM model when the LSTM network data prediction model is not updated; when the LSTM network data prediction model is updated, the LSTM network data prediction model acquired by the data prediction model acquisition module is the updated LSTM network data prediction model; and the learning rate adopted by the data prediction model updating module when the LSTM network data prediction model is trained is lower than the learning rate adopted by the LSTM network data prediction model when the LSTM network data prediction model is pre-trained.
CN202011500119.7A 2020-12-18 2020-12-18 MES system data online filling method and system based on LSTM network Pending CN112612782A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011500119.7A CN112612782A (en) 2020-12-18 2020-12-18 MES system data online filling method and system based on LSTM network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011500119.7A CN112612782A (en) 2020-12-18 2020-12-18 MES system data online filling method and system based on LSTM network

Publications (1)

Publication Number Publication Date
CN112612782A true CN112612782A (en) 2021-04-06

Family

ID=75240350

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011500119.7A Pending CN112612782A (en) 2020-12-18 2020-12-18 MES system data online filling method and system based on LSTM network

Country Status (1)

Country Link
CN (1) CN112612782A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114490596A (en) * 2021-12-08 2022-05-13 大唐水电科学技术研究院有限公司 Method for cleaning transformer oil chromatographic data based on machine learning and neural network

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114490596A (en) * 2021-12-08 2022-05-13 大唐水电科学技术研究院有限公司 Method for cleaning transformer oil chromatographic data based on machine learning and neural network
CN114490596B (en) * 2021-12-08 2024-05-10 大唐水电科学技术研究院有限公司 Method for cleaning transformer oil chromatographic data based on machine learning and neural network

Similar Documents

Publication Publication Date Title
CN111830408B (en) Motor fault diagnosis system and method based on edge calculation and deep learning
CN110739031B (en) Supervised prediction method and device for metallurgical sintering process and storage medium
CN115366157B (en) Industrial robot maintenance method and device
CN111030889B (en) Network traffic prediction method based on GRU model
CN113534741A (en) Control method and system for milling thin-walled workpiece
CN112612782A (en) MES system data online filling method and system based on LSTM network
CN114297912A (en) Tool wear prediction method based on deep learning
CN116451848A (en) Satellite telemetry data prediction method and device based on space-time attention mechanism
CN112651519A (en) Secondary equipment fault positioning method and system based on deep learning theory
CN114757087A (en) Tool wear prediction method based on dynamic principal component analysis and LSTM
CN113609766A (en) Soft measurement method based on depth probability latent model
CN115577647B (en) Power grid fault type identification method and intelligent agent construction method
CN114297795B (en) PR-Trans-based mechanical equipment residual life prediction method
CN114021469B (en) Method for monitoring one-stage furnace process based on mixed sequence network
CN115965177A (en) Improved autoregressive error compensation wind power prediction method based on attention mechanism
Luo et al. A novel method for remaining useful life prediction of roller bearings involving the discrepancy and similarity of degradation trajectories
CN114266286A (en) Online detection method and device for welding process information
CN114800049A (en) Grating ruler processing operation signal error compensation system
Aminnayeri et al. Short-run process control based on non-conformity degree
Zheng et al. Fault prediction of fan gearbox based on deep belief network
CN113009888B (en) Production line equipment state prediction and recognition device
CN117807410B (en) Method and device for determining set speed of steel-turning roller, storage medium and terminal
CN114580798B (en) Device point location prediction method and system based on transformer
CN112859793B (en) Industrial production process dynamic time delay identification method based on improved sliding time window
CN117592870B (en) Comprehensive analysis system based on water environment monitoring information

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20210406

RJ01 Rejection of invention patent application after publication