CN114925569A - Steel plate quenching temperature prediction method combining finite element and neural network - Google Patents
Steel plate quenching temperature prediction method combining finite element and neural network Download PDFInfo
- Publication number
- CN114925569A CN114925569A CN202210575794.9A CN202210575794A CN114925569A CN 114925569 A CN114925569 A CN 114925569A CN 202210575794 A CN202210575794 A CN 202210575794A CN 114925569 A CN114925569 A CN 114925569A
- Authority
- CN
- China
- Prior art keywords
- neural network
- steel plate
- training
- temperature
- layer
- 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
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
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- 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
- G06N3/04—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
- G06N3/08—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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Control Of Heat Treatment Processes (AREA)
Abstract
The invention discloses a steel plate quenching temperature prediction method combining a finite element model and a neural network, which comprises the steps of processing time temperature change curves of different positions of a steel plate in a quenching process calculated by the finite element model into training samples of the neural network, setting the number of nodes and training parameters of the neural network, establishing a neural network prediction model, dividing the training set and the verification set of the training samples, and obtaining a high-precision time temperature change curve prediction model of different positions of the steel plate in the quenching process through verification of the verification set after the neural network is trained, so as to solve the defect that a large number of working conditions cannot be predicted due to long calculation time of a quenching numerical model, thereby providing effective basis for making a proper quenching process and improving the performance of the steel plate in the heat treatment process.
Description
Technical Field
The invention belongs to the technical field of heat treatment, and particularly relates to a steel plate quenching temperature prediction method combining a finite element and a neural network.
Background
The heat treatment is used as a basic process for steel processing, is an indispensable part in the plate production process, and can obviously improve the mechanical property and the processing property of the steel plate after the heat treatment. In actual quenching, the temperature evolution law of the steel plate in the quenching process is mastered, so that the heat treatment process can be guided, and the quality of the heat-treated steel plate can be ensured. Therefore, the study of the temperature field in the quenching process is of great significance for optimizing the technological parameters of the cooling control equipment and establishing an accurate process control model.
In actual production, the steel plate quenching is mainly carried out through arranging complicated nozzle jet flow for cooling, the specification of each nozzle is different, the water supply pressure, the jet flow height and the jet flow angle of the nozzles are different according to the arrangement of a high-pressure section and a low-pressure section, and the cooling speed and the uniformity of the steel plate are difficult to control. At present, the temperature prediction of the quenched steel plate is mainly based on computational fluid dynamics, and the quenching process is numerically simulated by setting boundary conditions. A temperature field model is established through finite element software, the temperature distribution and evolution law of the steel plate can be obtained through simulation calculation, an accurate quenching steel plate temperature field needs to be obtained, all working conditions of actual quenching need to be considered, the corresponding heat exchange coefficient also needs to be close to the reality, the more real the boundary conditions adopted by the finite element quenching temperature field model are, the more the time calculated by the model is multiplied, and therefore, the time cost required for calculating and analyzing all working conditions in actual production is almost unacceptable.
The neural network has the characteristics of parallel processing, robustness, self-adaptability and self-learning, can approximate the characteristics of any nonlinear system, does not need to know the internal mechanism of the system, and can obtain a good prediction model only by adopting a corresponding network structure and training input and output data of the system. In recent years, the method is widely applied to a heat treatment heating steel plate temperature prediction model, but most of the methods only stay at a layer of predicting the furnace temperature, and a good steel temperature neural network model does not exist in the quenching process.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a steel plate quenching temperature prediction method combining a finite element and a neural network so as to solve the defect that a large number of working conditions cannot be predicted due to long calculation time of a quenching numerical model.
In order to achieve the purpose, the invention adopts the following technical scheme to realize the purpose:
a steel plate quenching temperature prediction method combining finite elements and a neural network comprises the following steps:
step 1, aiming at steel grades to be predicted, selecting partial working conditions according to thicknesses, and calculating a temperature field and temperature change curves of all positions through a finite element numerical model.
And 2, converting the temperature curve obtained in the step 1 into training data for input and output of the neural network by taking the thickness of the steel plate, the cooling time and the time interval of 0.5s from 0 as input parameters of the neural network, taking the temperature of the corresponding position of the steel plate corresponding to the current time as output parameters of the neural network and converting the temperature curve into training data for input and output of the neural network.
And 3, dividing the training data in the step 2 into a training set and a verification set.
And 4, carrying out normalization processing on the training set data.
And 5, determining the number of neuron nodes of each layer of the neural network, establishing a neural network model, and setting network parameters.
And 6, training the neural network model by using the training set in the step 4.
And 7, testing the prediction performance of the neural network by using the test set in the step 3.
The invention is further improved in that:
preferably, in step 3, the training set is 5/6 of the training data, and the validation set is 1/6 of the training data, that is, if there are 6 temperature curves, 5 temperature curves are the training set, and the remaining curve is the validation set.
Preferably, in step 4, the normalization process of the input and output data is to convert the training set by using a mapminmax function.
Preferably, in step 5, each layer of nodes of the neuron includes an input layer node, a hidden layer node, and an output layer node.
Preferably, in step 5, the node numbers of each layer of the neuron include an input layer node number of 3, a hidden layer node number of 3, and an output layer node number of 1.
Preferably, in step 5, the neural network parameters include: an input layer to hidden layer transfer function, a hidden layer to output layer transfer function, a training function, a learning function, a network iteration number, a training error target, a learning rate, and a minimum number of validation failures.
Preferably, in step 5, the neural network parameters are set as: the input layer to hidden layer transfer function is logsig, the hidden layer to output layer transfer function is tansig, the training function is train lm, the learning function leanngdm, the number of network iterations is 1000, the training error target is 1e-7, the learning rate is 0.01, and the minimum number of validation failures is 10.
Compared with the prior art, the invention has the following beneficial effects:
the invention discloses a steel plate quenching temperature prediction method combining a finite element model and a neural network, which takes a plate temperature time change curve obtained by simulating the finite element model as a training sample, establishes a 3-layer BP neural network model with a single hidden layer for different positions in the steel plate quenching process, predicts the temperature change of different positions in the steel plate quenching process, does not need to calculate the temperature of each steel plate through the finite element model, can predict the quenching temperature of a large number of working conditions only by training one part of the same type of steel through the neural network, and saves the time of calculating a large number of working conditions by the complex finite element model. According to the predicted time-dependent change curve of the temperature of the quenched steel plate, whether the quenching process is reasonable under the working condition is judged, corresponding optimization and improvement measures are carried out on the quenching process, data support can be provided for formulation of the subsequent quenching process, and parameters of the subsequent quenching process can be improved.
Drawings
FIG. 1 is a schematic flow chart of the present invention;
FIG. 2 is a selected 6 operating condition center cooling curves in the thickness range of 10-60mm in example 1 of the present invention;
FIG. 3 is a comparison of the prediction results of the neural network in example 1 of the present invention,
Detailed Description
The invention is described in further detail below with reference to the accompanying drawings:
referring to fig. 1, the invention provides a steel plate quenching temperature prediction method combining finite elements and a neural network, comprising the following steps:
step 1, selecting steel plate working conditions in a predicted range according to the thickness, calculating a temperature field of the selected working conditions and temperature change curves of all positions through a finite element numerical model, and extracting temperature time change curves of different positions.
And 2, converting the temperature curves of different positions obtained in the step 1 into training data input and output by a neural network according to the input and output parameters to form a data file, wherein the input parameters are the thickness of the steel plate, the cooling time and the time interval of 0.5s from 0, and the output parameters are the temperature of the corresponding position of the steel plate corresponding to the current time.
And 3, dividing the training data in the step 2 into a training set and a verification set. The training set is 5/6 for total data and the validation set is 1/6 for total data.
And 4, carrying out normalization processing on the training set data, wherein the adopted function is mapminmax.
Step 5, the neural network structure is determined to be a 3-layer BP neural network structure of a single hidden layer, and the node number of each layer of the neuron comprises: the number of nodes of the input layer is 3, the number of nodes of the hidden layer is 3, and the number of nodes of the output layer is 1. Establishing a neural network model, and setting neural network parameters as follows: the transfer function from the input layer to the hidden layer is logsig, the transfer function from the hidden layer to the output layer is tansig, the training function is train lm, the learning function learngdm, the network iteration number is 1000, the training error target is 1e-7, the learning rate is 0.01, and the minimum number of validation failures is 10.
And 6, training the neural network model by using the training set in the step 4, and repeatedly training until the set training target is reached.
And 7, testing the prediction accuracy of the neural network by using the test set in the step 3.
When the steel plate quenching temperature prediction method combining the finite elements and the neural network is used for establishing the model, the neural network prediction model in any working condition range can be established, and then the results are recalculated and analyzed in the steps 1-7.
The technical effects are as follows: the method can predict the temperature change of different positions in the quenching process of the steel plate, thereby providing a judgment basis for formulating a reasonable quenching process and achieving the purpose of optimizing the performance of the quenched steel plate.
The method for predicting the quenching temperature of the steel plate by combining the finite elements and the neural network is characterized in that the modeling software is MATLAB.
Examples
Referring to fig. 2, for NM450 steel plate, the initial cooling temperature is 1183K, and the thickness range is 10-60mm, the specific steps include:
1) according to the thickness range, working conditions of 10mm, 20mm, 30mm, 40mm, 50mm and 60mm are selected, and 6 working condition center cooling curves are calculated through a finite element model.
2) The input parameters are the thickness of the steel plate, the cooling time and the time interval of 0.5s from 0, the output parameters are the temperature of the center position of the steel plate corresponding to the current time, and 6 cooling temperature time change curves and corresponding input and output parameter data are converted into a data file in a table form.
3) And dividing the training data into a training set and a verification set. The training set was 5/6 and the validation set was 1/6 of the total data. Namely, 5 curves with the thickness of 10mm, 20mm, 40mm, 50mm and 60mm are taken as training sets, and a curve with the thickness of 30mm is taken as a verification set.
4) Normalizing the training set data, wherein the adopted function is mapminmax, specifically [ inputn, inputps ] ═ mapminmax (input _ train); [ outputn, outputps ] ═ mapminmax (output _ train);
5) the neural network structure is determined as a 3-layer BP neural network structure with a single hidden layer, and the node number of each layer of the neuron comprises: the number of nodes of the input layer is 3, the number of nodes of the hidden layer is 3, and the number of nodes of the output layer is 1. Establishing a neural network model, and setting neural network parameters as follows: the input layer to hidden layer transfer function is logsig, the hidden layer to output layer transfer function is tansig, the training function is rainlm, the learning function leanngdm, the number of network iterations is net, trainparam, epochs 1000, the training error target is net, trainparam, gold 1e-7, the learning rate is net, trainparam, lr 0.01, and the minimum number of acknowledgment failures is net, trainparam, max _ fail 10.
6) And training the neural network model by using the training set, and repeatedly training until the set training target is reached.
7) And testing the prediction accuracy of the neural network by using the test set.
FIG. 3 is a comparison of a test set temperature profile to a predicted temperature profile. It can be seen that the prediction performance of the neural network is better, and the maximum error at the low-voltage section is within 5K.
Claims (10)
1. A steel plate quenching temperature prediction method combining finite elements and a neural network is characterized by comprising the following steps:
step 1, selecting partial working conditions according to the thickness of steel grades to be predicted, and calculating a temperature field and temperature change curves of all positions through a finite element numerical model;
step 2, converting the temperature curve obtained in the step 1 into training data for input and output of the neural network by taking the thickness of the steel plate, the cooling time and the time interval of 0.5s from 0 as input parameters of the neural network, taking the temperature of the corresponding position of the steel plate corresponding to the current time as output parameters of the neural network;
step 3, dividing the training data in the step 2 into a training set and a verification set;
step 4, carrying out normalization processing on the training set data;
step 5, determining the number of neuron nodes of each layer of the neural network, establishing a neural network model, and setting network parameters;
step 6, training a neural network model by using the training set in the step 4;
and 7, testing the prediction performance of the neural network by using the test set in the step 3.
2. The method for predicting the quenching temperature of the steel plate with the combination of the finite element and the neural network as claimed in claim 1, wherein in the step 1, the part of the working conditions selected according to the thickness is the working conditions with the thickness of 5mm-10mm from small to large at intervals within the thickness range of the steel grade to be predicted, so that the full coverage of the thickness range is achieved.
3. The method for predicting quenching temperature of a steel plate by combining finite elements and a neural network as claimed in claim 1, wherein in the step 2, the temperature curve is converted into training data input and output by the neural network, and the data of the thickness of the steel plate, the cooling time, the time interval of 0.5s from 0 and the temperature of the corresponding position of the steel plate at the current time are converted into a table data file.
4. The method of claim 1, wherein in step 3, the training set is 5/6 of the training data in step 2, and the verification set is 1/6 of the training data, that is, if there are 6 temperature curves, 5 temperature curves are the training set, and the remaining curve is the verification set.
5. A method for predicting the quenching temperature of a steel plate with a finite element combined with a neural network as claimed in claim 1, wherein in step 4, the input and output data are normalized by converting the training set by using a mapminmax function.
6. A method for predicting quenching temperature of a steel plate with a finite element and a neural network combined as claimed in claim 1, wherein in step 5, each layer node of the neural network comprises an input layer node, a hidden layer node and an output layer node.
7. The method for predicting the quenching temperature of a steel plate with a finite element combined with a neural network as claimed in claim 6, wherein in the step 5, the node number of each layer of the neural element comprises an input layer node number of 3, a hidden layer node number of 3 and an output layer node number of 1.
8. The method for predicting the quenching temperature of the steel plate by combining the finite elements and the neural network as claimed in claim 1, wherein in the step 5, the neural network model is a 3-layer BP neural network with a single hidden layer.
9. The method for predicting quenching temperature of a steel plate by combining finite elements and a neural network according to claim 1, wherein in the step 5, the neural network parameters comprise: an input layer to hidden layer transfer function, a hidden layer to output layer transfer function, a training function, a learning function, a network iteration number, a training error target, a learning rate, and a minimum number of validation failures.
10. The method for predicting the quenching temperature of a steel plate with a finite element combined with a neural network as claimed in claim 9, wherein in the step 5, the neural network parameters are set as follows: the transfer function from the input layer to the hidden layer is logsig, the transfer function from the hidden layer to the output layer is tansig, the training function is train lm, the learning function learngdm, the network iteration number is 1000, the training error target is 1e-7, the learning rate is 0.01, and the minimum number of validation failures is 10.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210575794.9A CN114925569A (en) | 2022-05-25 | 2022-05-25 | Steel plate quenching temperature prediction method combining finite element and neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210575794.9A CN114925569A (en) | 2022-05-25 | 2022-05-25 | Steel plate quenching temperature prediction method combining finite element and neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114925569A true CN114925569A (en) | 2022-08-19 |
Family
ID=82810520
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210575794.9A Pending CN114925569A (en) | 2022-05-25 | 2022-05-25 | Steel plate quenching temperature prediction method combining finite element and neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114925569A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115563736A (en) * | 2022-10-28 | 2023-01-03 | 江南大学 | Turbine blade arc additive real-time temperature field prediction method |
-
2022
- 2022-05-25 CN CN202210575794.9A patent/CN114925569A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115563736A (en) * | 2022-10-28 | 2023-01-03 | 江南大学 | Turbine blade arc additive real-time temperature field prediction method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109583585B (en) | Construction method of power station boiler wall temperature prediction neural network model | |
CN112163380B (en) | Neural network hearth oxygen concentration prediction system and method based on numerical simulation | |
CN102880905B (en) | Online soft measurement method for normal oil dry point | |
CN104778298A (en) | Gaussian process regression soft measurement modeling method based on EGMM (Error Gaussian Mixture Model) | |
CN106779384B (en) | Iron and steel industry blast furnace gas long-term interval prediction method based on optimal information granularity distribution | |
CN104915518A (en) | Establishing method and application of two-dimensional prediction model of silicon content in hot metal in blast furnace | |
CN108388762A (en) | Sinter chemical composition prediction technique based on depth confidence network | |
CN109492335B (en) | Method and system for predicting furnace temperature of annealing furnace | |
CN114925569A (en) | Steel plate quenching temperature prediction method combining finite element and neural network | |
CN112016754A (en) | Power station boiler exhaust gas temperature advanced prediction system and method based on neural network | |
CN112884012A (en) | Building energy consumption prediction method based on support vector machine principle | |
CN115344019A (en) | Natural gas metering flow adjusting process based on composite intelligent algorithm | |
CN111931436A (en) | Burner nozzle air quantity prediction method based on numerical simulation and neural network | |
CN109918704B (en) | Die forging die life prediction method based on finite element simulation | |
CN103593550A (en) | Pierced billet quality modeling and prediction method based on integrated mean value staged RPLS-OS-ELM | |
CN111027258A (en) | Intelligent prediction method for generating load and heating load of supercritical unit | |
CN106990768A (en) | MKPCA batch process fault monitoring methods based on Limited DTW | |
CN108984943A (en) | Heating furnace steel billet temperature trace model modification method | |
CN101221437B (en) | Industrial production full process optimizing and controlling method in network information interchange mode | |
Yan et al. | Reliability prediction of CNC machine tool spindle based on optimized cascade feedforward neural network | |
CN113159395A (en) | Deep learning-based sewage treatment plant water inflow prediction method and system | |
CN110147645B (en) | Simulation model verification and establishment method and application in thin-wall copper pipe welding production process | |
CN116796406A (en) | Building energy consumption optimization method based on artificial neural network and BIM | |
CN111879910A (en) | Test method for optimizing forging process parameters and structure performance | |
CN116757078A (en) | Method and system for measuring flow velocity of pulverized coal based on acting force |
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 |