CN115204040A - Digital heating furnace simulation method based on neural network model - Google Patents

Digital heating furnace simulation method based on neural network model Download PDF

Info

Publication number
CN115204040A
CN115204040A CN202210718876.4A CN202210718876A CN115204040A CN 115204040 A CN115204040 A CN 115204040A CN 202210718876 A CN202210718876 A CN 202210718876A CN 115204040 A CN115204040 A CN 115204040A
Authority
CN
China
Prior art keywords
furnace
neural network
temperature
heating furnace
section
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
CN202210718876.4A
Other languages
Chinese (zh)
Inventor
于政军
王广胜
宋景威
李梅娟
陈雪波
董建新
柳昊彤
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.)
Dalian Xinruichen Automation Technology Co ltd
Original Assignee
Dalian Xinruichen Automation Technology Co ltd
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 Dalian Xinruichen Automation Technology Co ltd filed Critical Dalian Xinruichen Automation Technology Co ltd
Priority to CN202210718876.4A priority Critical patent/CN115204040A/en
Publication of CN115204040A publication Critical patent/CN115204040A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/10—Numerical modelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/08—Thermal analysis or thermal optimisation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computer Hardware Design (AREA)
  • Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • Geometry (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Control Of Heat Treatment Processes (AREA)

Abstract

The invention belongs to the technical field of billet heating furnace system simulation, and discloses a digital heating furnace simulation method based on a neural network model. And (3) taking the actual production data of the heating furnace as an input training sample and the furnace temperature as a prediction output by using a neural network to obtain a neural network model of the relation between the measurable variable in the heating furnace and the furnace temperature. And (3) predicting and simulating the actual furnace temperature of the heating furnace by using the furnace temperature of the neural network model, and combining the furnace temperature with a secondary intelligent control system to construct a digital heating furnace simulation platform. The digital heating furnace simulation platform is used for simulating the actual production environment of the heating furnace, the experimental environment related to the heating furnace system control algorithm is provided, the project development period is shortened, and the project implementation cost is saved.

Description

Digital heating furnace simulation method based on neural network model
Technical Field
The invention belongs to the technical field of billet heating furnace system simulation, and relates to a digital heating furnace simulation method based on a neural network model. In particular to a heating furnace simulation method of a heating furnace temperature forecasting model based on an artificial neural network.
Background
Because the heating furnace system is a nonlinear complex system with strong coupling, large time delay and multiple inputs and outputs, the research on the temperature control of the heating furnace is of great significance to the production of steel.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a digital heating furnace simulation method based on a neural network model. Therefore, the feasibility analysis of the designed control algorithm can be made before the actual project falls on the ground, so that the project development period is shortened, the time in field debugging is shortened, and the project cost is saved.
The above purpose of the invention is realized by the following technical scheme:
a digital heating furnace simulation method based on a neural network model is characterized in that the neural network model is used for training data of measurable variables produced in the actual production of a heating furnace as input samples, and the furnace temperature is used as a prediction output result; the input samples are the temperature of the furnace section, the temperature of the previous furnace section, the temperature of the next furnace section, the gas flow of the furnace section, the air flow of the furnace section, the total volume of the steel billet of the furnace section and the heat content of the steel billet of the furnace section;
the heating furnace is a multivariable influence nonlinear system which is mutually coupled, and an LSTM neural network is adopted to model the temperature of the heating furnace; because the heating furnace has 7 heating sections, one heating section is subjected to neural network modeling;
the mathematical model relationship of the furnace temperature and other variables is as follows:
Figure BDA0003710574520000021
Figure BDA0003710574520000022
the parameters in formula (1.1) are: t is z (k) The furnace temperature of an output result predicted by the neural network of the furnace section at the time k is shown, T is the temperature (DEG C), z is the furnace section of the heating furnace, z belongs to {1,2,3,4,5,6,7}, k is a certain discrete time,
Figure BDA0003710574520000023
the same applies below; t is z (k-1)、T z-1 (k-1)、T z+1 (k-1) represents the furnace temperature at the moment of k-1 of the present furnace section, the previous furnace section and the next furnace section; g z (k-1) represents the gas flow rate at the time k-1; a. The z (k-1) represents the air flow rate at time k;
Figure BDA0003710574520000024
represents the heat content of the total billet in the furnace section at the moment k-1,
Figure BDA0003710574520000025
Figure BDA0003710574520000026
Figure BDA0003710574520000027
denotes the volume, T, of each billet n Is a temperature representation of each billet;
Figure BDA0003710574520000028
shows the total volume of the steel billet in the furnace section at the time k,
Figure BDA0003710574520000029
Figure BDA00037105745200000210
representing the volume of each billet;
f(·) Is a non-linear function, and
Figure BDA00037105745200000211
n u ，n k for the order of system input and output, R is a set of real numbers, L 2 (R n ) Is the Hilbert space.
And establishing a furnace temperature prediction model by using an LSTM neural network.
The training data samples of the LSTM neural network are the data of the variables in equation (1.1) at time k.
The inputs to the neural network are:
Figure BDA00037105745200000212
the output of the network is: t is a unit of z (k)。
The activation function selects the ReLU function, which is a non-linear activation function.
The gradient optimization algorithm adopts an Adam algorithm. The formula of the Adam algorithm is shown below: m is t ＝μm t-1 +(1-β 1 )g t ，
Figure BDA00037105745200000213
m t Is an estimate of the first moment of the gradient; n is t Is an estimate of the second moment of the gradient; μ, v ∈ [0, 1), exponential decay rate for moment estimation;
Figure BDA00037105745200000214
is to m t Correcting;
Figure BDA00037105745200000215
is to n t Correcting; eta is learning rate, delta theta t To update the parameters, g t For time step gradients, ε is the error rate.
And (4) preprocessing data. Firstly, preprocessing is carried out on data, and detection and filtering of abnormal values are mainly carried out. Because a plurality of variable characteristic attributes exist in the heating furnace data, the maximum value and the minimum value of each variable are different, the model learning capacity can be increased to a certain extent by adding normalization when the model is constructed, the model precision is improved, and the convergence speed is accelerated.
Linear scaling of data sets to the range [0,1 ] using linear transformations]In the interior of said container body,
Figure BDA0003710574520000031
in the formula x min And x max The minimum and maximum values of the data set, respectively. The main advantage of scaling is to avoid that properties in a larger range of values dominate properties in a smaller range of values.
The number of hidden layer layers in the neural network is selected to be 2, the number of neurons in each hidden layer is 380, the learning rate is 0.01, the number of training iterations of the neural network is 500, and the Adam algorithm is selected by the gradient descent optimization algorithm.
And (6) performance evaluation. There are several methods that can be used to measure the accuracy of the time series prediction model. For this type of prediction, the prediction accuracy is checked by calculating three commonly used evaluation indices: root Mean Square Error (RMSE), mean Absolute Percent Error (MAPE), and Mean Absolute Error (MAE). The expressions for MAPE, RMSE and MAE are:
Figure BDA0003710574520000032
wherein
Figure BDA0003710574520000033
And y t Respectively a predicted value and an actual value; k is the number of predictions.
And judging that RMSE is less than 0.6, MAPE is less than 0.05 and MAE is less than 0.5, and successfully training the model.
Compared with the prior art, the invention has the beneficial effects that:
and (3) taking the actual production data of the heating furnace as an input training sample and the furnace temperature as a prediction output by using the neural network to obtain a neural network model of the relation between the measurable variable in the heating furnace and the furnace temperature. The actual furnace temperature of the heating furnace is predicted and simulated by using the neural network model, and a digital heating furnace simulation platform is constructed by combining with a traditional two-stage intelligent control system of a steel mill. The simulation platform of the digital heating furnace is utilized to simulate the actual production environment of the heating furnace, the experimental environment related to the control algorithm of the heating furnace system is provided, the project development period is shortened, and the project implementation cost is saved.
Drawings
The invention is further illustrated with reference to the following figures and examples:
FIG. 1 is a schematic diagram of an LSTM neural network furnace temperature prediction model.
FIG. 2 is a chart of a 1-stage LSTM-NN furnace temperature training fit in a heating furnace.
FIG. 3 is a graph of the predicted furnace temperature of the LSTM-NN heating section 1 in the heating furnace.
FIG. 4 is a block diagram of a heating furnace simulation system based on a neural network.
FIG. 5 is a graph of the actual temperature of the furnace versus the predicted temperature for example 1.
Detailed Description
The invention is described in more detail below with reference to specific examples, without limiting the scope of the invention. Unless otherwise specified, the experimental methods adopted by the invention are all conventional methods, and experimental equipment, materials, reagents and the like used in the experimental method can be obtained from commercial sources.
Example 1
As shown in fig. 1-4, a neural network-based digital heating furnace simulation method is provided, which trains the data of measurable variables actually produced by the heating furnace as input samples by using a neural network model, and takes the furnace temperature as a prediction output result. The input samples are the temperature of the furnace section, the temperature of the previous furnace section, the temperature of the next furnace section, the gas flow of the furnace section, the air flow of the furnace section, the total volume of the steel billet of the furnace section and the heat content of the steel billet of the furnace section.
The furnace temperature data of the heating furnace is collected according to a time sequence, and the furnace temperature of the heating furnace is predicted by adopting an LSTM neural network. The training data is actual field data and can almost cover the main working condition of field change. And (3) taking the furnace temperature in the annular heating furnace as a prediction output result, and taking the temperature of the upper and lower sections, the temperature of the local section, the gas flow, the air flow, the steel billet volume of the section, the steel billet temperature of the section and the like as input data to perform model training. The model training is carried out by using a PyTorch framework, the display card adopts GeForce RTX 2060, and the CPU is intel Core i7-9700F. 20000 groups of data are totally collected in the training set, the sampling period of the data is 10s, the test data volume is 360 pieces of data, and the verification set is 360 pieces of data.
The heating furnace is a multivariable influence and mutually coupled nonlinear system, and an LSTM neural network is adopted to model the temperature of the heating furnace. Since there are 7 heating zones in the furnace, a neural network modeling is now performed for one of the heating zones.
The mathematical model relationship of the furnace temperature and other variables is as follows:
Figure BDA0003710574520000051
Figure BDA0003710574520000052
the parameters in formula (1.1) are: t is z (k) Showing the furnace temperature of the output result predicted by the neural network of the furnace section at the time k, T showing the temperature (DEG C), z showing the furnace section of the heating furnace, z belonging to {1,2,3,4,5,6,7}, k showing a certain discrete time,
Figure BDA0003710574520000053
the same applies below; t is z (k-1)、T z-1 (k-1)、T z+1 (k-1) represents the furnace temperature at the moment of k-1 of the present furnace section, the previous furnace section and the next furnace section; g z (k-1) represents the gas flow rate at the time k-1; a. The z (k-1) represents the air flow rate at time k;
Figure BDA0003710574520000054
represents the heat content of the total billet in the furnace section at the moment k-1,
Figure BDA0003710574520000055
Figure BDA0003710574520000056
Figure BDA0003710574520000057
denotes the volume, T, of each billet n Is a temperature representation of each billet;
Figure BDA0003710574520000058
shows the total volume of the steel billet in the furnace section at the time k,
Figure BDA0003710574520000059
Figure BDA00037105745200000510
representing the volume of each billet;
f (-) is a non-linear function, and
Figure BDA00037105745200000511
n u ，n k for the order of system input and output, R is a set of real numbers, L 2 (R n ) Is the Hilbert space.
And establishing a furnace temperature prediction model by using an LSTM neural network.
The training data samples of the LSTM neural network are the data of the variables in equation (1.1) at time k.
The inputs to the neural network are:
Figure BDA00037105745200000512
the output of the network is: t is z (k)。
The activation function selects the ReLU function, which is a non-linear activation function.
The gradient optimization algorithm adopts an Adam algorithm. The formula of the Adam algorithm is shown below: m is t ＝μm t-1 +(1-μ)g t ，
Figure BDA0003710574520000061
m t Is an estimate of the first moment of the gradient; n is a radical of an alkyl radical t Is an estimate of the second moment of the gradient; μ, v ∈ [0, 1), exponential decay rate for moment estimation;
Figure BDA0003710574520000062
is to m t Correcting;
Figure BDA0003710574520000063
is to n t Correcting; eta is learning rate, delta theta t To update the parameters, g t ε is the error rate, the gradient of the time step.
And (4) preprocessing data. Firstly, preprocessing is carried out on data, and detection and filtering of abnormal values are mainly carried out. Because the heating furnace data has a plurality of variable characteristic attributes, the maximum value and the minimum value of each variable are different, the model learning capacity can be increased to a certain extent by adding normalization when the model is constructed, the model precision is improved, and the convergence speed is accelerated.
Linear scaling of data sets to the range [0,1 ] using linear transformations]In the interior of the container body,
Figure BDA0003710574520000064
in the formula x min And x max Respectively, the minimum and maximum values of the data set. The main advantage of scaling is to avoid that properties in a larger range of values dominate properties in a smaller range of values.
Selecting the number of hidden layer layers of the neural network as 2 layers, the number of neurons in each hidden layer as 380, the learning rate as 0.01, the training iteration times of the neural network as 500, and selecting an Adam algorithm by a gradient descent optimizing algorithm.
And (6) performance evaluation. There are several methods that can be used to measure the accuracy of the time series prediction model. For this type of prediction, the prediction accuracy is checked by calculating three commonly used evaluation indices: root mean square error (, RMSE), mean Absolute Percent Error (MAPE), and Mean Absolute Error (MAE). The expressions for MAPE, RMSE and MAE are:
Figure BDA0003710574520000065
wherein
Figure BDA0003710574520000066
And y t Respectively a predicted value and an actual value; k is the number of predictions.
And judging that RMSE is less than 0.6, MAPE is less than 0.05 and MAE is less than 0.5, and successfully training the model.
The experimental results are as follows: fig. 2 is a curve fitted to the heating 1 stage furnace temperature training process. It can be seen that the predicted furnace temperature fits well to the actual value. FIG. 3 is a diagram of furnace temperature prediction of heating 1 segment LSTM-NN (NN is an abbreviation of Neural Network), and the number of predicted samples is 360. The blue curve is an actual value, the red curve is a predicted value of the furnace temperature, and it can be seen that each predicted value curve has no particularly large discrete degree with the actual value curve, and the predicted value is almost close to a real value.
According to the furnace temperature prediction accuracy verification data of the LSTM-NN heating section 1, the minimum value of the root mean square error RMSE is 0.5551, the minimum value of the mean absolute error MAE is 0.4429, and the minimum value of the mean absolute error MAPR is 0.035, which shows that the predicted value has small error with the actual value, the accuracy is high, and the predicted value accords with the expected furnace temperature prediction result.
In the furnace temperature prediction of the heating 1 section, the real furnace temperature curve is in an oscillation phenomenon, LSTM-NN can also be well predicted to follow, and an overfitting phenomenon does not exist. In the furnace temperature prediction of the heating sections 2-4, the middle steep peak is in a temperature rising state, and the LSTM-NN can also accurately predict according to the trend of the curve. In conclusion, the LSTM-NN can achieve a good prediction result on the furnace temperature.
The working principle and the using process of the invention are as follows: FIG. 4 is a block diagram of a heating furnace simulation system based on a neural network, and the heating furnace secondary intelligent control system is firstly installed, a production plan is recorded, and the secondary intelligent control system is operated. The secondary intelligent control system can monitor the zone furnace temperature, the zone billet number, the zone billet total volume, the zone billet total heat content, the gas flow, the air flow, the running speed and the like of the heating furnace, and store the data into a database in real time according to the sampling time of 1 s. A steel temperature field model calculation module is arranged in the secondary intelligent control system, and the temperature of the steel billet is calculated in real time.
The neural network furnace temperature prediction model after training is loaded on the basis of the Pythroch development platform, data such as the temperature of the furnace section, the temperature of the previous furnace section, the temperature of the next furnace section, the gas flow of the furnace section, the air flow of the furnace section, the total volume of steel billets of the furnace section, the heat content of the steel billets of the furnace section and the like in the database are accessed and used as input of the neural network model, the furnace temperature is predicted, and the predicted furnace temperature is stored in the database. And the secondary intelligent control system reads the furnace temperature value predicted by the neural network model in the database in real time and displays the furnace temperature value in a picture.
When a steel billet with larger length or diameter enters the furnace section in a single section in the secondary intelligent control system, the volume and the total heat content value of the steel billet are increased, the sampling data at the moment are stored in a database through the secondary intelligent control system, the data in the database are read out by a neural network model, and the predicted furnace temperature is reduced, thereby conforming to the actual expectation of the furnace temperature rise of the heating furnace. When steel billets with smaller length or diameter enter the furnace section in a single section in the secondary intelligent control system, the volume and the total heat content of the steel billets become smaller, the sampled data at the moment are stored in a database through the secondary intelligent control system, the data in the database are read out by the neural network model, the predicted furnace temperature becomes larger, and the furnace temperature reduction expectation of the heating furnace in practice is met.
When the operation speed of the secondary intelligent control system is increased, the heating time of heating steel billets in the heating furnace simulation system is shortened, the acceleration of the steel temperature rise is reduced, the total heat content of the steel billets in the section is increased slowly, the sampled data is stored in the database, and after the data in the database is read through the neural network model, the predicted furnace temperature is increased compared with the predicted furnace temperature at the previous moment, so that the furnace temperature increase expectation of the actual heating furnace is met. The neural network model takes data sampled by the secondary intelligent control system as input, and the predicted furnace temperature result accords with the actual field working condition of the heating furnace.
When the gas flow and the air flow are increased in the two-stage intelligent control system, the section furnace temperature rises: when the gas flow and the air flow are reduced in the secondary system, the section furnace temperature is reduced, and the predicted furnace temperature result accords with the actual site working condition of the heating furnace.
Application example 1
The method provided in example 1 is applied to the furnace temperature prediction of a heating furnace in a certain steel mill in Jiangsu. As shown in fig. 5, a graph of the actual furnace temperature of the furnace versus the predicted furnace temperature is shown. The curve of the predicted furnace temperature in the graph is very good to fit with the curve of the actual furnace temperature, and the temperature difference between the predicted furnace temperature and the actual furnace temperature is less than or equal to +/-1 ℃, so that a good prediction effect is achieved.
The embodiments described above are merely preferred embodiments of the invention, rather than all possible embodiments of the invention. Any obvious modifications to the above would be obvious to those of ordinary skill in the art, but would not bring the invention so modified beyond the spirit and scope of the present invention.

Claims (5)

1. A digital heating furnace simulation method based on a neural network model is characterized in that the neural network model is utilized to train measurable variable data of actual production of a heating furnace as an input sample, and the furnace temperature is used as a prediction output result; the input samples are the temperature of the furnace section, the temperature of the previous furnace section, the temperature of the next furnace section, the gas flow of the furnace section, the air flow of the furnace section, the total volume of the steel billet of the furnace section and the heat content of the steel billet of the furnace section.
2. The simulation method of the digital heating furnace based on the neural network model as claimed in claim 1, wherein the neural network model is used for training measurable variable data of actual production of the heating furnace as an input sample, and the furnace temperature is used as a prediction output result; the input samples are the temperature of the furnace section, the temperature of the previous furnace section, the temperature of the next furnace section, the gas flow of the furnace section, the air flow of the furnace section, the total volume of the steel billet of the furnace section and the heat content of the steel billet of the furnace section;
the heating furnace is a multivariable influence and mutually coupled nonlinear system, and an LSTM neural network is adopted to model the temperature of the heating furnace; because the heating furnace has 7 heating sections, one heating section is subjected to neural network modeling;
the mathematical model relationship of the furnace temperature and other variables is as follows:
Figure FDA0003710574510000011
Figure FDA0003710574510000012
the parameter in the formula (1.1) is：T z (k) Showing the furnace temperature of the output result predicted by the neural network of the furnace section at the time k, T showing the temperature (DEG C), z showing the furnace section of the heating furnace, z belonging to {1,2,3,4,5,6,7}, k showing a certain discrete time,
Figure FDA0003710574510000013
the same applies below; t is z (k-1)、T z-1 (k-1)、T z+1 (k-1) represents the furnace temperature at the moment of k-1 of the present furnace section, the previous furnace section and the next furnace section; g z (k-1) represents the gas flow rate at the time k-1; a. The z (k-1) represents the air flow rate at time k;
Figure FDA0003710574510000014
represents the heat content of the total billet in the furnace section at the moment k-1,
Figure FDA0003710574510000015
Figure FDA0003710574510000016
Figure FDA0003710574510000017
denotes the volume, T, of each billet n The temperature of each billet is represented;
Figure FDA0003710574510000018
shows the total volume of the steel billet in the furnace section at the time k,
Figure FDA0003710574510000019
Figure FDA00037105745100000110
representing the volume of each billet;
f (-) is a non-linear function, and
Figure FDA00037105745100000111
n u ，n k is input into the systemOrder of output, R is a set of real numbers, L 2 (R n ) Is a Hilbert space;
establishing a furnace temperature prediction model by using an LSTM neural network;
the training data samples of the LSTM neural network are the data of the variables in equation (1.1) at time k.
The inputs to the neural network are:
Figure FDA0003710574510000021
the output of the network is: t is z (k)；
The activation function selects a ReLU function;
the gradient optimization algorithm adopts an Adam algorithm;
preprocessing data;
and (5) performance evaluation.
3. The method for simulating the digital heating furnace based on the neural network model as claimed in claim 2, wherein the formula of the Adam algorithm is as follows: m is a unit of t ＝μm t-1 +(1-β 1 )g t ，
Figure FDA0003710574510000022
Figure FDA0003710574510000023
m t Is an estimate of the first moment of the gradient; n is t Is an estimate of the second moment of the gradient; μ, v ∈ [0, 1), exponential decay rate for moment estimation;
Figure FDA0003710574510000024
is to m t Correcting;
Figure FDA0003710574510000025
is to n t Correcting; eta is learning rate, delta theta t To update the parameters, g t ε is the error rate, the gradient of the time step.
4. The simulation method of the digital heating furnace based on the neural network model as claimed in claim 3, wherein the data preprocessing comprises: firstly, preprocessing data, mainly comprising abnormal value detection and filtering; linear scaling of data sets to the range [0,1 ] using a linear transformation]In the interior of said container body,
Figure FDA0003710574510000026
in the formula x min And x max Respectively, a minimum value and a maximum value of the data set;
the number of hidden layer layers in the neural network is selected to be 2, the number of neurons in each hidden layer is 380, the learning rate is 0.01, the number of training iterations of the neural network is 500, and the Adam algorithm is selected by the gradient descent optimization algorithm.
5. The simulation method of the digital heating furnace based on the neural network model as set forth in claim 4, wherein the performance evaluation is to check the prediction accuracy by calculating three commonly used evaluation indexes: RMSE, MAPE and MAE; the expressions for MAPE, RMSE and MAE are:
Figure FDA0003710574510000031
Figure FDA0003710574510000032
wherein
Figure FDA0003710574510000033
And y t Respectively a predicted value and an actual value; k is the number of predictions; and judging that RMSE is less than 0.6, MAPE is less than 0.05 and MAE is less than 0.5, and successfully training the model.
CN202210718876.4A 2022-06-23 2022-06-23 Digital heating furnace simulation method based on neural network model Pending CN115204040A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210718876.4A CN115204040A (en) 2022-06-23 2022-06-23 Digital heating furnace simulation method based on neural network model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210718876.4A CN115204040A (en) 2022-06-23 2022-06-23 Digital heating furnace simulation method based on neural network model

Publications (1)

Publication Number Publication Date
CN115204040A true CN115204040A (en) 2022-10-18

Family

ID=83577767

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210718876.4A Pending CN115204040A (en) 2022-06-23 2022-06-23 Digital heating furnace simulation method based on neural network model

Country Status (1)

Country Link
CN (1) CN115204040A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117057212A (en) * 2023-06-05 2023-11-14 四川大学 Acoustic reconstruction method for dynamic temperature field of nuclear power device

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117057212A (en) * 2023-06-05 2023-11-14 四川大学 Acoustic reconstruction method for dynamic temperature field of nuclear power device
CN117057212B (en) * 2023-06-05 2024-05-17 四川大学 Acoustic reconstruction method for dynamic temperature field of nuclear power device

Similar Documents

Publication Publication Date Title
CN108764517B (en) Method, equipment and storage medium for predicting change trend of silicon content in molten iron of blast furnace
CN107725283B (en) A kind of fan trouble detection method based on deepness belief network model
CN114282443B (en) Residual service life prediction method based on MLP-LSTM supervised joint model
WO2020147461A1 (en) Data driven modeling and optimization method for device for deasphalting by solvent
CN113325721B (en) Model-free adaptive control method and system for industrial system
CN114015825B (en) Method for monitoring abnormal state of blast furnace heat load based on attention mechanism
CN108897354B (en) Aluminum smelting process hearth temperature prediction method based on deep belief network
CN112381279B (en) Wind power prediction method based on VMD and BLS combined model
CN116303786B (en) Block chain financial big data management system based on multidimensional data fusion algorithm
CN115034129B (en) NOx emission concentration soft measurement method for thermal power plant denitration device
CN114239397A (en) Soft measurement modeling method based on dynamic feature extraction and local weighted deep learning
CN116432699A (en) Method for predicting silicon content of blast furnace molten iron based on time sequence chart packing network
CN111159863A (en) Neural network algorithm-based material deformation and failure prediction method and system
CN115204040A (en) Digital heating furnace simulation method based on neural network model
CN104050380A (en) LF furnace final temperature forecasting method based on Adaboost-PLS-ELM
Yin et al. A double layer BRB model for health prognostics in complex electromechanical system
CN106842948A (en) The method for optimally controlling of the HDP molecular distillation systems based on BP networks
CN112784331A (en) Soil stress-strain relation determination method based on improved LSTM deep learning method
CN115062528A (en) Prediction method for industrial process time sequence data
CN114036827A (en) Multi-target carbon emission reduction method for blast furnace ironmaking based on decomposition
CN117578441A (en) Method for improving power grid load prediction precision based on neural network
CN117034808A (en) Natural gas pipe network pressure estimation method based on graph attention network
CN111625995A (en) Online time-space modeling method integrating forgetting mechanism and double ultralimit learning machines
JP2009076036A (en) Process state prediction method
Wang et al. Concrete compression test data estimation based on a wavelet neural network model

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