CN100520786C - Processing parameter setting method of tension straightening machine set correct roller - Google Patents
Processing parameter setting method of tension straightening machine set correct roller Download PDFInfo
- Publication number
- CN100520786C CN100520786C CNB2006100281797A CN200610028179A CN100520786C CN 100520786 C CN100520786 C CN 100520786C CN B2006100281797 A CNB2006100281797 A CN B2006100281797A CN 200610028179 A CN200610028179 A CN 200610028179A CN 100520786 C CN100520786 C CN 100520786C
- Authority
- CN
- China
- Prior art keywords
- parameter
- network
- output
- layer
- straightening machine
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B21—MECHANICAL METAL-WORKING WITHOUT ESSENTIALLY REMOVING MATERIAL; PUNCHING METAL
- B21D—WORKING OR PROCESSING OF SHEET METAL OR METAL TUBES, RODS OR PROFILES WITHOUT ESSENTIALLY REMOVING MATERIAL; PUNCHING METAL
- B21D1/00—Straightening, restoring form or removing local distortions of sheet metal or specific articles made therefrom; Stretching sheet metal combined with rolling
- B21D1/02—Straightening, restoring form or removing local distortions of sheet metal or specific articles made therefrom; Stretching sheet metal combined with rolling by rollers
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Straightening Metal Sheet-Like Bodies (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a kind of setting method for processing parameter of correct roller of withdrawal straightening stands based on spot measured data and analysis result, and the shape of steel sheet at the export of withdrawal straightening stands is good as standard, and it chooses the relative data with representation as the training sample parameter of artificial nerve network, and combines the characteristics of withdrawal straightening stands to choose the skilled and general reversing transmission nerve network. The good operation experience is converted into mathematical model to build the setting method for processing parameter under different steel rule and material. The invention can overcome the problems that the setting of processing parameter of traditional spot withdraw is short of theory, and the result is not ideal, and the invention fits for the production characteristics, it has good withdraw effect and satisfy the requirement for producing the production with high quality.
Description
Technical field
The present invention relates to the rectification of the cold-reduced sheet plate shape of cold rolling of strip steel aftertreatment, more specifically be meant a kind of processing parameter setting method of tension straightening machine set correct roller.
Background technology
For cold rolling aftertreatment unit, move back unit as galvanizing and company, the common straightening machine that is provided with behind planisher is to improve the strip shape quality of band steel, and the structure of general straightening machine sees also shown in Figure 1, and the straightening machine is made of 1# bending roll group 11,2# bending roll group 12, vertical strong roller 13 and horizontal strong roller 14.Under certain extensibility,, form differently curved stretching, thereby improve belt plate shape the band steel by setting the difference insertion amount of these rollers.
At present, the processing parameter setting value of these rollers of straightening machine unit generally is an empirical data, mostly is the developer greatly with reference value that its knowhow provided.In actual production, the operator generally relies on experience separately to carry out the setting of straightening machine.This plate shape improvement that is limited by the situation of different order of classes or grades at school operative employee experiences not only bad for the straightening of band steel, nor be beneficial to the ordinary production of unit and expansion of product from now on and exploitation, cause the setting of on-the-spot straightening technological parameter to lack the corresponding theory guidance, effect is undesirable, is difficult to satisfy the demand of produced on-site high quality of products.
Therefore, at existing problem in the present straightening machine parameter setting process, processing parameter setting method that must a kind of new tension straightening machine set correct roller of design can be fit to unit self production products characteristics, have straightening effect preferably, can satisfy the demand of produced on-site high quality of products.
Summary of the invention
The object of the present invention is to provide a kind of processing parameter setting method of new tension straightening machine set correct roller, lack the corresponding theory guidance to overcome conventional on-site straightening processing parameter setting, the unfavorable problem of effect, can be fit to unit self production products characteristics, have straightening effect preferably, can satisfy the demand of produced on-site high quality of products.
To achieve these goals, the present invention has adopted following technical scheme:
The processing parameter setting method of this tension straightening machine set correct roller mainly comes the parameter of correct roller is set based on the multilayer feedforward neural network structure of back propagation learning mode, specifically may further comprise the steps:
A. artificial neural network's training sample selection of parameter is determined, the related data of the steel plate that selection will be corrected, comprise shape wave data, band steel matter, band steel geometric parameter, the wide parameter of plate and flattening technological parameter and the straightening extensibility of being with steel, as artificial neural network's training sample parameter, wherein: described shape wave data comprise transmission side limit portion steepness, transmission side 1/4 place's steepness, middle part steepness, fore side 1/4 place's steepness and operation side portion steepness; Described band steel matter is meant yield limit; Described band steel geometric parameter is meant width, the thickness of band steel; The wide parameter of described plate comprises secondary convexity, four convexitys; Described flattening technological parameter comprises smooth extensibility, unit roll-force, bending roller force;
B. model structure determines, in conjunction with the withdrawal straightening stands own characteristic, selected the neural network of three layers of back propagation learning algorithm that have only a hidden layer for use, the input layer unit number is determined by unit access panel shape influence factor and flattening technological parameter, the output of network is determined according to the setup parameter of straightening machine, hidden layer node rule of thumb formula is determined, wherein: described model structure determine to select for use three layers of backpropagation neuroid of 14-12-4 types structure, promptly, described input layer unit number is defined as 14, described hidden layer node number is defined as 12, and the output node number of described network is defined as 4;
C. the computation process of counterpropagation network may further comprise the steps:
C1. with one group of sample parameter input neuron;
C2. input parameter is passed to hidden layer through weighted by input layer;
C3. import weighting parameters and after function activates, obtain the hidden layer output valve;
C4. the hidden layer output valve passes after weighted to output layer;
C5. the hidden layer weighting parameters obtains the network output valve after the output layer function activates, i.e. the setup parameter of straightening machine;
Adopt formula in the computation process of described counterpropagation network:
M=14 wherein, x
1, x
2..., x
mBe input signal, for input layer, be 14 node parameter values that unit access panel shape influence factor and flattening technological parameter are determined; w
K1, w
K2..., w
Kn1The cynapse that is neuron k connects weights, the random number between its initial value desirable (1,1); u
kBe the output of the linear combiner of input signal,
Be activation function, y
kBe the neuron output signal, for output layer, be 4 setup parameters of straightening machine;
D. the learning process of counterpropagation network may further comprise the steps:
D1. network output valve and target output value are relatively drawn the output layer error, calculate the mean square deviation of sample;
D2. with error back propagation, the cynapse of each interlayer of roll-off network successively is connected weights with the gradient descent method to utilize mean square deviation E, and error is constantly reduced, and this makeover process is:
Gradient descent method by mean square deviation E gets
The correction algorithm that cynapse connects weights is w
Kj(k
+ 1)=w
Kj(k)+η δ
j(k) y
k(k)
In the formula, η is a learning rate, δ
j(k) be error term, W
Kj(k) be the cynapse connection weights of neuron k, y
k(k) be the neuron output signal;
For accelerating training speed, introduce momentum term, then the correction algorithm of cynapse connection weights is
w
kj(k+1)＝w
kj(k)+ηδ
j(k)y
k(k)+α[w
kj(k)-w
kj(k-1)]
In the formula, α is a factor of momentum;
D3. repeating step C finishes until learning sample;
D4. repeating step C until mean square deviation less than expectation value.
In technique scheme of the present invention, this method is based on field measurement data statistics and analysis result, and well be standard with straightening machine unit outlet belt plate shape quality, therefrom choose the training sample of representational related data according to qualifications as the artificial neural network, in conjunction with the withdrawal straightening stands own characteristic, choose backpropagation neuroid structure comparatively ripe and commonly used in artificial neural network's instrument.The most on-the-spot good operating experience is converted into mathematical model, thus set up straightening technological parameter under different steel specification, the material condition establishing method.The present invention can overcome conventional on-site straightening processing parameter setting and lack the corresponding theory guidance, and the unfavorable problem of effect is fit to unit self production products characteristics, has straightening effect preferably, can satisfy the demand of produced on-site high quality of products.
Description of drawings
Fig. 1 is the structural representation of traditional straightening machine.
Fig. 2 is the processing parameter setting method schematic flow sheet of tension straightening machine set correct roller of the present invention.
Fig. 3 is a straightening machine counterpropagation network structural representation.
Fig. 4 is neuronic nonlinear model.
Fig. 5 is a counterpropagation network mean square deviation change curve.
Fig. 6 sets front and rear panel shape comparison diagram for the neuroid straightening.
Embodiment
In order to understand technique scheme of the present utility model better, describe in detail further below in conjunction with drawings and Examples.
See also shown in Figure 2ly, the processing parameter setting method of tension straightening machine set correct roller of the present invention mainly comes the parameter of correct roller is set based on the multilayer feedforward neural network structure of back propagation learning mode.In fact method of the present invention is to have selected BP neuroid structure comparatively ripe and commonly used in artificial neural network's instrument for use, and so-called BP network refers to the multilayer feedforward neural network that adopts backpropagation (Back Propagation) learning algorithm.
Method of the present invention specifies as follows:
At first, artificial neural network's training sample selection of parameter is determined, the related data of the steel plate that selection will be corrected, comprise shape wave data, band steel matter, band steel geometric parameter, the wide parameter of plate and flattening technological parameter and the straightening extensibility of being with steel, as artificial neural network's training sample parameter.
Described shape wave data comprise transmission side limit portion steepness, transmission side 1/4 place's steepness, middle part steepness, fore side 1/4 place's steepness and operation side portion steepness;
Described band steel matter is meant yield limit, and described band steel geometric specifications is meant width, the thickness of band steel;
The wide parameter of described plate comprises secondary convexity, four convexitys, and described flattening technological parameter comprises smooth extensibility, unit roll-force, bending roller force.
Then, determining of model structure, choose backpropagation neuroid structure comparatively ripe and commonly used in artificial neural network's instrument, in conjunction with the withdrawal straightening stands own characteristic, the backpropagation neuroid has here adopted the 3-tier architecture network model, promptly has only three layers of counterpropagation network of a hidden layer.See also shown in Figure 3, the input layer unit number is determined by unit access panel shape influence factor and flattening technological parameter, input layer amounts to 14 node numbers, the shape wave data that comprised the band steel, as 5 steepness: transmission side portion steepness 201, transmission side 1/4 place's steepness 202, middle part steepness 203, fore side 1/4 place's steepness 204 and operation side portion steepness 205, the yield limit 206 of band steel matter, strip width 207, thickness 208, the wide parameter of plate is as secondary convexity 209, four convexitys 210 and smooth smooth extensibility 211, unit roll-force 212, bending roller force 213 and straightening extensibility 214; Network is output as 4 setup parameters of straightening machine, comprises 1# roller insertion depth 31,2# roller insertion depth 32, vertical roller insertion amount 33 and horizontal roller insertion amount 34, so the node number of output is 4; Hidden layer node rule of thumb formula gets 12, and the final backpropagation neuroid structure of determining is 14-12-4.
Then, carry out the computation process of counterpropagation network, passed to hidden layer through weighted by input layer as withdrawal straightening stands access panel shape influence factor and 14 definite node parameter values of flattening technological parameter, function obtains the hidden layer output valve after activating; The hidden layer output valve passes after weighted to output layer, obtains the network output valve after the output layer function activates, i.e. 4 of the straightening machine setup parameters.
When carrying out the computation process of counterpropagation network, undertaken by following step:
C1. with one group of sample parameter input neuron,
C2. input parameter is passed to hidden layer through weighted by input layer,
C3. import weighting parameters and after function activates, obtain the hidden layer output valve,
C4. the hidden layer output valve passes after weighted to output layer,
C5. the hidden layer weighting parameters obtains the network output valve after the output layer function activates.
For counterpropagation network, the artificial neuron models of one of them typical case can adopt down the description that establishes an equation as shown in Figure 4:
M=14 wherein, x
1, x
2..., x
mBe input signal, for input layer, be 14 node parameter values that unit access panel shape influence factor and flattening technological parameter are determined; w
K1, w
K2..., w
KmThe cynapse that is neuron k connects weights, the random number between its initial value desirable (1,1); u
kBe the output of the linear combiner of input signal,
Be activation function, y
kBe the neuron output signal, for output layer, be 4 setup parameters of straightening machine.
At last, carry out the study (promptly revising) of counterpropagation network again, output valve and target output value are relatively obtained error, with error back propagation, and successively the cynapse of each interlayer of roll-off network connects weights, and error is constantly reduced.Repeat above training process, till error satisfies accuracy requirement.
Particularly, learning process further may further comprise the steps:
D1. draw the mean square deviation of output layer Error Calculation sample,
D2. the cynapse of revising each layer of network connects weights,
D3. repeating step C finishes until learning sample,
D4. repeating step C until mean square deviation less than expectation value.
The learning process of counterpropagation network is exactly to utilize mean square deviation E and gradient descent method to realize that to network cynapse is connected the correction of weights.
Gradient descent method by mean square deviation E gets
The correction algorithm that cynapse connects weights is w
Kj(k+1)=w
Kj(k)+η δ
j(k) y
k(k)
In the formula, η is a learning rate, δ
j(k) be error term.
For accelerating training speed, introduce momentum term, then the correction algorithm of cynapse connection weights is
w
kj(k+1)＝w
kj(k)+ηδ
j(k)y
k(k)+α[w
kj(k)-w
kj(k-1)]
In the formula, α is a factor of momentum.
Concrete setting of counterpropagation network parameter and value are:
Frequency of training between twice demonstration=100; Learning rate=0.05; The learning rate growth ratio factor=1.05; Learning rate decline scale factor=0.8; Frequency of training=4000; Network training target=0.08; Mean square deviation expectation value=0.001; Factor of momentum=0.9.
The mean square deviation change curve of counterpropagation network learning process as shown in Figure 5, wherein horizontal ordinate is a frequency of training, ordinate is a mean square deviation.By change curve as can be known, mean square deviation is along with the increase of frequency of training reduces successively.Therefore, error is that the increase along with frequency of training reduces gradually, and is more satisfactory.See also the situation of adjustment front and back belt steel plate deformationization shown in Figure 6.Behind the processing parameter setting method that adopts new tension straightening machine set correct roller, plate shape has obtained improving significantly.
By foregoing description as seen, the present invention constantly adjusts the connection weights of network by error back propagation, make full use of counterpropagation network and have good self study, self-adaptation, height Nonlinear Mapping, dynamic fault-tolerant, advantage such as efficient, realize the purpose of straightening machine parameter setting.The present invention can overcome conventional on-site straightening processing parameter setting and lack the corresponding theory guidance, and the unfavorable problem of effect is fit to unit self production products characteristics, has straightening effect preferably, can satisfy the demand of produced on-site high quality of products.
Those of ordinary skill in the art will be appreciated that, above embodiment is used for illustrating the present invention, and be not to be used as limitation of the invention, as long as in connotation scope of the present invention, all will drop in claims scope of the present invention variation, the modification of the above embodiment.
Claims (2)
1, a kind of processing parameter setting method of tension straightening machine set correct roller,
It is characterized in that,
This method comes the parameter of correct roller is set based on the multilayer feedforward neural network structure of back propagation learning mode, specifically may further comprise the steps:
A. artificial neural network's training sample selection of parameter is determined, the related data of the steel plate that selection will be corrected, comprise shape wave data, band steel matter, band steel geometric parameter, the wide parameter of plate and flattening technological parameter and the straightening extensibility of being with steel, training sample parameter as the artificial neural network, wherein, described shape wave data comprise transmission side limit portion steepness, transmission side 1/4 place's steepness, middle part steepness, fore side 1/4 place's steepness and operation side portion steepness; Described band steel matter is meant yield limit; Described band steel geometric parameter is meant width, the thickness of band steel; The wide parameter of described plate comprises secondary convexity, four convexitys; Described flattening technological parameter comprises smooth extensibility, unit roll-force, bending roller force;
B. model structure determines, in conjunction with the withdrawal straightening stands own characteristic, selected the neural network of three layers of back propagation learning algorithm that have only a hidden layer for use, the input layer unit number is determined by unit access panel shape influence factor and flattening technological parameter, the output of network is determined according to the setup parameter of straightening machine, hidden layer node rule of thumb formula is determined, wherein, described model structure determine to select for use three layers of backpropagation neuroid of 14-12-4 types structure, promptly, described input layer unit number is defined as 14, and described hidden layer node number is defined as 12, and the output node number of described network is defined as 4;
C. the computation process of counterpropagation network may further comprise the steps,
C1. with one group of sample parameter input neuron;
C2. input parameter is passed to hidden layer through weighted by input layer;
C3. import weighting parameters and after function activates, obtain the hidden layer output valve;
C4. the hidden layer output valve passes after weighted to output layer;
C5. the hidden layer weighting parameters obtains the network output valve after the output layer function activates, i.e. the setup parameter of straightening machine;
Adopt formula in the computation process of described counterpropagation network:
M=14 wherein, x
1, x
2..., x
mBe input signal, for input layer, be 14 node parameter values that unit access panel shape influence factor and flattening technological parameter are determined;
The cynapse that is neuron k connects weights, the random number between its initial value desirable (1,1); u
kBe the output of the linear combiner of input signal,
Be activation function, y
kBe the neuron output signal, for output layer, be 4 setup parameters of straightening machine;
D. the learning process of counterpropagation network may further comprise the steps:
D1. network output valve and target output value are relatively drawn the mean square deviation of output layer Error Calculation sample;
D2. with error back propagation, the cynapse of each interlayer of roll-off network successively is connected weights with the gradient descent method to utilize mean square deviation E, and error is constantly reduced, and this makeover process is:
Gradient descent method by mean square deviation E gets
The correction algorithm that cynapse connects weights is w
Kj(k+1)=w
Kj(k)+η δ
j(k) y
k(k)
In the formula, η is a learning rate, δ
j(k) be error term, W
Kj(k) be the cynapse connection weights of neuron k, y
k(k) be the neuron output signal;
For accelerating training speed, introduce momentum term, then the correction algorithm of cynapse connection weights is
w
kj(k+1)＝w
kj(k)+ηδ
j(k)y
k(k)+α[w
kj(k)-w
kj(k-1)]
In the formula, α is a factor of momentum;
D3. repeating step C finishes until learning sample;
D4. repeating step C until mean square deviation less than expectation value.
2, the processing parameter setting method of tension straightening machine set correct roller as claimed in claim 1,
It is characterized in that,
Counterpropagation network parameter setting and value were when the cynapse of each interlayer of error back propagation roll-off network connected weights among the described step D2:
Learning rate=0.05; The learning rate growth ratio factor=1.05; Learning rate decline scale factor=0.8; Frequency of training=4000; Network training target=0.08; Mean square deviation expectation value=0.001; Factor of momentum=0.9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2006100281797A CN100520786C (en) | 2006-06-27 | 2006-06-27 | Processing parameter setting method of tension straightening machine set correct roller |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2006100281797A CN100520786C (en) | 2006-06-27 | 2006-06-27 | Processing parameter setting method of tension straightening machine set correct roller |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101097581A CN101097581A (en) | 2008-01-02 |
| CN100520786C true CN100520786C (en) | 2009-07-29 |
Family
ID=39011413
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2006100281797A Active CN100520786C (en) | 2006-06-27 | 2006-06-27 | Processing parameter setting method of tension straightening machine set correct roller |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100520786C (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102042893B (en) * | 2009-10-23 | 2013-04-24 | 宝山钢铁股份有限公司 | Soft-measuring method for tension of band steel between rollers of continuous annealing unit |
| CN102672003B (en) * | 2011-03-07 | 2015-04-01 | 宁波宝新不锈钢有限公司 | Method for setting process parameters of stainless steel strip steel withdrawal and straightening machine unit |
| CN102601166A (en) * | 2012-03-05 | 2012-07-25 | 上海海事大学 | Straightening method based on support vector machine for workpiece deformation |
| CN104376226A (en) * | 2014-11-28 | 2015-02-25 | 首钢总公司 | Method and system for calculating band steel thickness reduction amount |
| CN105404711B (en) * | 2015-10-23 | 2018-10-19 | 首钢集团有限公司 | A kind of optimization method and system of withdrawal straightening machine technological parameter |
| CN106862284B (en) * | 2017-03-24 | 2018-09-04 | 燕山大学 | A kind of cold rolled sheet signal mode knowledge method for distinguishing |
| CN107808241B (en) * | 2017-10-16 | 2021-08-06 | 山西太钢不锈钢股份有限公司 | Stainless steel surface detection result comprehensive analysis system |
| CN108306997B (en) * | 2018-01-25 | 2021-03-23 | 中国工商银行股份有限公司 | Domain name resolution monitoring method and device |
| CN109821930B (en) * | 2019-02-14 | 2020-06-26 | 中国重型机械研究院股份公司 | Method for setting working parameters of stretch bending straightener |
| CN111814219B (en) * | 2020-06-30 | 2024-04-30 | 中冶京诚工程技术有限公司 | Parameterized design method and system for withdrawal and straightening machine of continuous casting machine |
-
2006
- 2006-06-27 CN CNB2006100281797A patent/CN100520786C/en active Active
Non-Patent Citations (8)
| Title |
|---|
| 人工神经网络在板料拉深成形中的应用研究. 王春良，陈明和.航空精密制造技术，第39卷第5期. 2003 |
| 人工神经网络在板料拉深成形中的应用研究. 王春良，陈明和.航空精密制造技术，第39卷第5期. 2003 * |
| 基于神经网络模式识别的板形模糊控制器. 刘建昌，王柱.东北大学学报(自然科学版)，第26卷第8期. 2005 |
| 基于神经网络模式识别的板形模糊控制器. 刘建昌，王柱.东北大学学报(自然科学版)，第26卷第8期. 2005 * |
| 基于遗传算法-反向传播模型的板形模式识别. 张材，谭建平.中南大学学报(自然科学版)，第37卷第2期. 2006 |
| 基于遗传算法-反向传播模型的板形模式识别. 张材，谭建平.中南大学学报(自然科学版)，第37卷第2期. 2006 * |
| 拉弯矫直机组工艺参数的确定. 陈秀琴.重型机械科技，第1期. 2005 |
| 拉弯矫直机组工艺参数的确定. 陈秀琴.重型机械科技，第1期. 2005 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101097581A (en) | 2008-01-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100520786C (en) | Processing parameter setting method of tension straightening machine set correct roller | |
| CN108637020A (en) | A kind of TSP question PSO-BP neural networks strip profile prediction technique | |
| CN101934290B (en) | Load allocation adjusting method for stainless steel tandem cold rolling mill | |
| WO2023130666A1 (en) | Strip steel plate convexity prediction method based on data-driving and mechanism model fusion | |
| CN101376139B (en) | Control method for producing conical plate blank using side compression machine of fixed width plate blank | |
| CN101513647B (en) | Method for leveling strip produced by secondary cold rolling unit | |
| CN102672003B (en) | Method for setting process parameters of stainless steel strip steel withdrawal and straightening machine unit | |
| CN108655186B (en) | Roll-force presetting method based on artificial neural network and mathematical model | |
| CN104942019B (en) | A kind of cold rolling of strip steel process Automatic control method of width | |
| CN102921743B (en) | Method for determining five stand tandem cold mill depressing distribution | |
| CN105740467A (en) | Mining method for C-Mn steel industry big data | |
| CN103745101A (en) | Improved neural network algorithm based forecasting method of set value of rolling force of medium plate | |
| CN103886373A (en) | High-speed cold-rolling mill third octave flutter prediction method based on BP neural network | |
| CN102294364A (en) | Method for presetting rolling force of extremely-thin board temper mill | |
| CN109440515A (en) | Soft measurement method for beating degree in papermaking process based on gradient enhanced regression tree | |
| CN102426439A (en) | Pierced billet quality forecasting and control method based on data driving | |
| CN111291513B (en) | BP neural network-based prediction method for rolling force of differential thick plate | |
| CN103506404A (en) | Method for controlling roll gap during finish rolling of strip steel | |
| CN103357669A (en) | Plate model prediction control method | |
| CN101648208A (en) | Rolling procedure making method of aluminium single-stand cold-rolling machine | |
| CN113373295A (en) | Annealing furnace digital twin intelligent alarm system and method based on multilayer perceptron | |
| CN107716560B (en) | A kind of Hot Strip Rolling load distribution method | |
| CN115007658B (en) | Cold continuous rolling strip shape dynamic prediction method based on deep circulation neural network | |
| CN101934288B (en) | Cold continuous rolling reduction distribution method | |
| CN106269911A (en) | A kind of roughing pressure Control System of Load Distribution method and roughing control system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |