CN112487700A - Cold rolling force prediction method based on NSGA and FELM - Google Patents
Cold rolling force prediction method based on NSGA and FELM Download PDFInfo
- Publication number
- CN112487700A CN112487700A CN202010963972.6A CN202010963972A CN112487700A CN 112487700 A CN112487700 A CN 112487700A CN 202010963972 A CN202010963972 A CN 202010963972A CN 112487700 A CN112487700 A CN 112487700A
- Authority
- CN
- China
- Prior art keywords
- felm
- individuals
- rolling force
- model
- nsga
- 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.)
- Granted
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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- 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
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Business, Economics & Management (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Economics (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Computing Systems (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Evolutionary Biology (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Operations Research (AREA)
- Physiology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Manufacturing & Machinery (AREA)
- Primary Health Care (AREA)
Abstract
The invention relates to a cold rolling force prediction method based on NSGA and FELM, belonging to the technical field of automatic control of a designed rolling process, and comprising the following steps: step 1, collecting an original production data sample of the strip steel, step 2, carrying out normalization processing on data in the sample, step 3, setting FELM network parameters, primarily establishing a FELM cold rolling force prediction model, step 4, optimizing related parameters of the FELM model by adopting a fast non-dominated genetic sorting algorithm NSGA-II, and carrying out model test. The invention realizes the prediction of the cold rolling force, overcomes the defects of lower precision and poor anti-interference capability of the traditional rolling force model, has high precision, can be put into use only by analyzing and processing a large amount of original production data and programming through a computer, and has lower cost.
Description
Technical Field
The invention relates to the technical field of automatic control of a rolling process, in particular to a cold rolling force prediction method based on NSGA and FELM.
Background
In the rolling process, the rolling force model cannot completely and accurately describe the rolling process because the on-site rolling conditions and the incoming material conditions are changed continuously. Conventional rolling models make a large number of assumptions during the derivation process, and thus there is often a difference between the set value and the actual value. Particularly, in the production process of cold continuous rolling strip steel, the quality problem of a first coil product after specification switching is always the primary problem troubling manufacturers, the rolling condition and state change in the specification switching process is large, the traditional rolling force mechanism model is difficult to realize accurate prediction of the rolling force of the first coil steel, and a method capable of improving the rolling force prediction precision in the specification changing process is urgently needed.
Disclosure of Invention
The invention aims to provide a cold rolling force prediction method based on NSGA and FELM, which realizes cold rolling force prediction according to actually measured rolling data in a rolling process and improves the effect of the rolling force prediction precision in a specification changing process;
in order to achieve the purpose, the invention adopts the following technical scheme:
a cold rolling force prediction method based on NSGA and FELM comprises the following steps:
step 1: the method comprises the following steps of collecting an original production data sample of the strip steel, wherein the sample data comprises: the width of the strip steel is 750mm-1000mm, the inlet thickness is 1.710mm-3.414mm, the outlet thickness is 1.093mm-2.463mm, the reduction rate is 21.6-40%, the inlet tension is 90.8kN-194kN, the outlet tension is 113.6kN-303.6kN, the inlet unit tension is 60.6MPa-77.4MPa, the outlet unit tension is 122.7MPa-161.7MPa, the diameter of the roller is 385.0mm-425.2mm, the rolling length is 0km-149.6km, the rolling speed is 75.4m/min-364.4m/min, and the rolling force is 512kN-862 kN;
step 2: carrying out normalization processing on data in the samples, and dividing the collected strip steel original production data samples into a training set, a verification set and a test set; the normalization processing steps are as follows:
step 2.1: normalizing the data in the sample, wherein the normalization calculation formula is as follows:
wherein y ismin，ymaxIs the default value: -1 or 1, since x is set to an initial value, y is the value after normalization, xmax，xminAre the maximum and minimum in the data set;
step 2.2: dividing a sample into a training set, a verification set and a test set, wherein the selection ratio of each set is 3: 1;
and step 3: setting the parameters of an FELM network, and primarily establishing an FELM cold rolling force prediction model, wherein the FELM is a feedback limit learning machine:
step 3.1: randomly selecting a connection weight IW between an input layer and a hidden layer of the FELM network and a threshold B of a hidden layer neuron, setting the number n of the initial hidden layer neurons as 100, and determining a hidden layer neuron activation function TF;
step 3.2: setting the TYPE parameter in the FELM to be 0, setting the initial feedback error to be 0, and setting the initial iteration number N to be 15;
step 3.3: starting iterative training network to obtain a current FELM model;
step 3.4: testing the obtained FELM model error by using a test set test network and recording a numerical value, correcting the number n of the neurons of the hidden layer according to the numerical value, and according to the step 3.3 and the step 3.4, until the number n of the neurons of the optimal hidden layer is found to be 150 in the value interval of 100 plus 200;
and 4, step 4: optimizing related parameters of the FELM model by adopting a fast non-dominated genetic sorting algorithm NSGA-II, and performing model test as follows:
step 4.1: encoding the optimized parameters before the genetic algorithm starts, and converting the parameters into a digital string consisting of 0 and 1 by adopting binary coding;
step 4.2: generating a population of k individuals, each individual of the population representing a set of solutions corresponding to the optimization objective function;
step 4.3: converting the target value of the rolling force prediction error of the strip steel into a corresponding fitness function, sequentially calculating the fitness of each individual in the population according to the fitness function, and sequencing the fitness from large to small to provide parameters for the subsequent evolution selection of the population; the fitness function calculation formula is as follows:
wherein, yiAnd
respectively a true value and a predicted value of the verification set, and p is the number of samples in the verification set;
step 4.4: judging whether the algorithm can be stopped, and stopping the operation if the algorithm meets the condition;
step 4.5: firstly, placing non-dominant individuals in all individuals into a set, wherein one set is a first Pareto plane, removing the individuals from all the individuals, and repeatedly placing the non-dominant individuals in all the individuals into one set again to find a next Pareto plane until all the individuals are classified into different Pareto planes; calculating crowding degree, namely the sum of target difference values between two adjacent individuals on the same Pareto plane and a target individual, namely the side length of a small rectangle formed by the two adjacent individuals, wherein each individual in a group is positioned on different Pareto planes and the individuals positioned on the same plane have different crowding degrees through rapid non-dominated sorting and crowding degree calculation, and if the two individuals are positioned on different Pareto planes, the individuals are separated to the Pareto planes firstly by adopting layered sorting; if two individuals are on the same Pareto plane, selecting individuals who are not crowded around;
step 4.6: selecting individuals from the population, and using the individuals as parents of the next generation to breed offspring so as to generate new individuals;
step 4.7: randomly selecting a certain gene site of parents for propagation, and mutually exchanging chromosomes of the parents after the chromosomes of the parents are broken at the site;
step 4.8: selecting a plurality of individuals from the population, and randomly selecting a certain position in the gene of the selected individual according to Gaussian distribution and replacing the selected individual with a random number;
step 4.9: and combining the genetic algorithm with the FELM to optimize the connection weight IW between the initial input layer and the hidden layer of the FELM and the threshold B of the neuron of the hidden layer so as to achieve the prediction precision of the model.
The technical scheme of the invention is further improved as follows: and 3.3, starting iterative training of the network, and finishing iterative training when the number of iterations is 15 to obtain the current FELM model.
The technical scheme of the invention is further improved as follows: in step 3.3, iterative training of the network is started, and when the error E of the verification setvLess than training set error ErAnd obtaining the current FELM model.
The technical scheme of the invention is further improved as follows: step 4.4: and judging whether the algorithm can be stopped, and stopping the operation of the algorithm when the maximum evolution algebra is set to be 100 and the population is evolved to the maximum algebra.
The technical scheme of the invention is further improved as follows: step 4.4: and judging whether the algorithm can be stopped, and stopping the operation when the stopping tolerance is set to be 0.001, namely the fitness function value reaches 0.001.
Compared with the prior art, the cold rolling force prediction method based on NSGA and FELM provided by the invention has the following beneficial effects:
1. the invention provides a cold rolling force prediction method based on NSGA and FELM, which utilizes an FELM network and combines with an NSGA-II algorithm to realize cold rolling force prediction and overcomes the defects of low precision and poor interference resistance of a traditional rolling force model.
2. The invention provides a cold rolling force prediction method based on NSGA and FELM, which has high precision, can be put into use only by analyzing and processing a large amount of original production data and programming through a computer, and has lower cost.
Drawings
FIG. 1 is a flow chart of a cold rolling force prediction method based on NSGA and FELM of the present invention;
FIG. 2 is a model predicted value and measured value scatter-point fitting graph of the cold rolling force prediction method based on NSGA and FELM.
Detailed Description
The technical solution of the present invention will be clearly and completely described by the following detailed description. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1-2, the method for predicting cold rolling force based on NSGA and FELM includes the following steps:
step 1: the method comprises the following steps of collecting an original production data sample of the strip steel, wherein the sample data comprises: the width of the strip steel is 750mm-1000mm, the inlet thickness is 1.710mm-3.414mm, the outlet thickness is 1.093mm-2.463mm, the reduction is 21.6-40%, the inlet tension is 90.8kN-194kN, the outlet tension is 113.6kN-303.6kN, the inlet unit tension is 60.6MPa-77.4MPa, the outlet unit tension is 122.7MPa-161.7MPa, the diameter of the roller is 385.0mm-425.2mm, the rolling length is 0km-149.6km, the rolling speed is 75.4m/min-364.4m/min, and the rolling force of the steel coil is 512kN-862kN, in the embodiment, the original production data sample of the strip steel is collected, and 888 steel coil samples are counted; each set of data comprises the 12 sets of samples;
step 2: carrying out normalization processing on data in the samples, and dividing the collected strip steel original production data samples into a training set, a verification set and a test set; the normalization processing steps are as follows:
step 2.1: normalizing the data in the sample, wherein the normalization calculation formula is as follows:
wherein y ismin，ymaxIs the default value: -1 or 1, since x is set to an initial value, y is the value after normalization, xmax，xminAre the maximum and minimum in the data set;
step 2.2: the method comprises the following steps of dividing a sample into a training set, a verification set and a test set, wherein the sorting proportion of each set is 3:1, in the embodiment, dividing data into the training set, the verification set and the test set, the selection principle is random selection, finally 700 steel coil production data are selected as the training set, 94 data are selected as the verification set, 94 data are selected as the test set, the training set accounts for about 80%, and the verification set and the test set respectively account for about 10%;
and step 3: setting an FELM network parameter, and primarily establishing an FELM cold rolling force prediction model, wherein the FELM is a feedback limit learning machine;
step 3.1: randomly selecting a connection weight IW between an input layer and a hidden layer of the FELM network and a threshold B of a hidden layer neuron, setting the number n of the initial hidden layer neurons as 100, and determining a hidden layer neuron activation function TF;
step 3.2: setting the TYPE parameter in the FELM to be 0, setting the initial feedback error to be 0, and setting the initial iteration number N to be 15;
step 3.3: starting iterative training network to obtain a current FELM model;
preferably, in step 3.3, the iterative training network is started, and when the number of iterations is 15, the iterative training is ended, so as to obtain the current FELM model;
preferably, in step 3.3, iterative training of the network is started, when the validation set error EvLess than training set error ErObtaining the currentA FELM model;
step 3.4: testing the obtained FELM model error by using a test set test network and recording a numerical value, correcting the number n of the neurons of the hidden layer according to the numerical value, and according to the step 3.3 and the step 3.4, until the number n of the neurons of the optimal hidden layer is found to be 150 in the value interval of 100 plus 200;
and 4, step 4: optimizing related parameters of the FELM model by adopting a fast non-dominated genetic sorting algorithm NSGA-II, and performing model test as follows:
step 4.1: encoding the optimized parameters before the genetic algorithm starts, and converting the parameters into a digital string consisting of 0 and 1 by adopting binary coding;
step 4.2: generating a population of k individuals, each individual of the population representing a set of solutions corresponding to the optimization objective function;
step 4.3: the target value of the rolling force prediction error of the strip steel is converted into a corresponding fitness function, the fitness of each individual in the population is sequentially calculated according to the fitness function, and the fitness is ranked from large to small, so that parameters are provided for the subsequent evolution selection of the population. The fitness function calculation formula is as follows:
wherein, yiAnd
respectively a true value and a predicted value of the verification set, and p is the number of samples in the verification set;
step 4.4: judging whether the algorithm can be stopped, and stopping the operation if the algorithm meets the condition;
preferably, step 4.4: judging whether the algorithm can be stopped, and stopping the operation of the algorithm when the maximum evolution algebra is set to be 100 and the population is evolved to the maximum algebra;
preferably, step 4.4: judging whether the algorithm can be stopped, and stopping the operation of the algorithm when the stopping tolerance is set to be 0.001, namely the fitness function value reaches 0.001;
step 4.5: firstly, placing non-dominant individuals in all individuals into a set, wherein one set is a first Pareto plane, removing the individuals from all the individuals, and repeatedly placing the non-dominant individuals in all the individuals into one set again to find a next Pareto plane until all the individuals are classified into different Pareto planes; calculating the crowding degree, namely the sum of target difference values between two adjacent individuals on the same Pareto plane and a target individual, namely the side length of a small rectangle formed by the two adjacent individuals, wherein each individual in the group is positioned on different Pareto planes and the individuals positioned on the same plane have different crowding degrees through rapid non-dominated sorting and crowding degree calculation, and if the two individuals are positioned on different Pareto planes, the individuals are separated to the Pareto planes firstly by adopting hierarchical sorting; if two individuals are on the same Pareto plane, selecting individuals who are not crowded around;
step 4.6: selecting individuals from the population, and using the individuals as parents of the next generation to breed offspring so as to generate new individuals; in the embodiment, a championship selection method is adopted, so that the probability of selecting an individual with a good shape is high, the good characters in the individual can be inherited to the next generation, but the individual with a common character also has the probability of being selected, and the diversity of the population is ensured to a certain extent; testing the conditions of 20, 30, 40 and 50 population scales, sequentially traversing and searching, and determining the optimal population scale to be 30;
step 4.7: randomly selecting a certain gene site of parents for propagation, and mutually exchanging chromosomes of the parents after the chromosomes of the parents are broken at the site; in this embodiment, a single-point crossing mode is adopted, that is, one point location is selected, and the left and right parts of the genes of parents are exchanged; the cross probability generally selects a larger value, the value is increased by 0.1 each time, and the 4 conditions of 0.6-0.9 are tested, and the test result shows that when the cross probability is 0.6, the performance of the model is optimal;
step 4.8: selecting a plurality of individuals from the population, and randomly selecting a certain position in the gene of the selected individual according to Gaussian distribution and replacing the selected individual with a random number;
as shown in fig. 2, step 4.9: combining a genetic algorithm with the FELM, optimizing a connection weight IW between an initial input layer and a hidden layer of the FELM and a threshold B of a neuron of the hidden layer, and improving the prediction precision of the model; in the embodiment, the NSGA-II algorithm and the FELM algorithm are combined, and the final performance of the model is tested by using the training set data, so that the error value is ensured to be smaller as much as possible; according to the prediction model provided by the invention, in the prediction values, almost all the steel coil rolling force prediction values are within a 5% error band, most of the steel coil rolling force prediction values are positioned near the error band, only a few abnormal values exist, most of the errors are very small, the mean square error value is 708.2614kN, and the average error percentage is only 3.0023%.
The above-mentioned embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements of the technical solution of the present invention by those skilled in the art should fall within the protection scope defined by the claims of the present invention without departing from the spirit of the present invention.
Claims (5)
1. A cold rolling force prediction method based on NSGA and FELM is characterized by comprising the following steps:
step 1: the method comprises the following steps of collecting an original production data sample of the strip steel, wherein the sample data comprises: the width of the strip steel is 750mm-1000mm, the inlet thickness is 1.710mm-3.414mm, the outlet thickness is 1.093mm-2.463mm, the reduction rate is 21.6-40%, the inlet tension is 90.8kN-194kN, the outlet tension is 113.6kN-303.6kN, the inlet unit tension is 60.6MPa-77.4MPa, the outlet unit tension is 122.7MPa-161.7MPa, the diameter of the roller is 385.0mm-425.2mm, the rolling length is 0km-149.6km, the rolling speed is 75.4m/min-364.4m/min, and the rolling force is 512kN-862 kN;
step 2: carrying out normalization processing on data in the samples, and dividing the collected strip steel original production data samples into a training set, a verification set and a test set; the sample data processing steps are as follows:
step 2.1: normalizing the data in the sample, wherein the normalization calculation formula is as follows:
wherein, ymin，ymaxIs the default value: -1 or 1, since x is set to an initial value, y is the value after normalization, xmax，xminAre the maximum and minimum in the data set;
step 2.2: dividing a sample into a training set, a verification set and a test set, wherein the selection proportion of each set is 3:1: 1;
and step 3: setting the parameters of an FELM network, and primarily establishing an FELM cold rolling force prediction model, wherein the FELM is a feedback limit learning machine:
step 3.1: randomly selecting a connection weight IW between an input layer and a hidden layer of the FELM network and a threshold B of a hidden layer neuron, setting the number n of the initial hidden layer neurons as 100, and determining a hidden layer neuron activation function TF;
step 3.2: setting the TYPE parameter in the FELM to be 0, setting the initial feedback error to be 0, and setting the initial iteration number N to be 15;
step 3.3: starting iterative training network to obtain a current FELM model;
step 3.4: testing the obtained FELM model error by using a test set test network and recording a numerical value, correcting the number n of the neurons of the hidden layer according to the numerical value, and according to the step 3.3 and the step 3.4, until the number n of the neurons of the optimal hidden layer is found to be 150 in the value interval of 100 plus 200;
and 4, step 4: optimizing related parameters of the FELM model by adopting a fast non-dominated genetic sorting algorithm NSGA-II, and performing model test steps as follows:
step 4.1: encoding the optimized parameters before the genetic algorithm starts, and converting the parameters into a digital string consisting of 0 and 1 by adopting binary coding;
step 4.2: generating a population of k individuals, each individual of the population representing a set of solutions corresponding to the optimization objective function;
step 4.3: converting the target value of the rolling force prediction error of the strip steel into a corresponding fitness function, sequentially calculating the fitness of each individual in the population according to the fitness function, and sequencing the fitness from large to small to provide parameters for the subsequent evolution selection of the population; the fitness function calculation formula is as follows:
wherein, yiAnd
respectively a true value and a predicted value of the verification set, and p is the number of samples in the verification set;
step 4.4: judging whether the algorithm can be stopped, and stopping the operation if the algorithm meets the condition;
step 4.5: firstly, placing non-dominant individuals in all individuals into a set, wherein one set is a first Pareto plane, removing the individuals from all the individuals, and repeatedly placing the non-dominant individuals in all the individuals into one set again to find a next Pareto plane until all the individuals are classified into different Pareto planes; calculating crowding degree, namely the sum of target difference values between two adjacent individuals on the same Pareto plane and a target individual, namely the side length of a small rectangle formed by the two adjacent individuals, wherein each individual in a group is positioned on different Pareto planes and the individuals positioned on the same plane have different crowding degrees through rapid non-dominated sorting and crowding degree calculation, and if the two individuals are positioned on different Pareto planes, the individuals are separated to the Pareto planes firstly by adopting layered sorting; if two individuals are on the same Pareto plane, selecting individuals who are not crowded around;
step 4.6: selecting individuals from the population, and using the individuals as parents of the next generation to breed offspring so as to generate new individuals;
step 4.7: randomly selecting a certain gene site of parents for propagation, and mutually exchanging chromosomes of the parents after the chromosomes of the parents are broken at the site;
step 4.8: selecting a plurality of individuals from the population, and randomly selecting a certain position in the gene of the selected individual according to Gaussian distribution and replacing the selected individual with a random number;
step 4.9: and combining the genetic algorithm with the FELM to optimize the connection weight IW between the initial input layer and the hidden layer of the FELM and the threshold B of the neuron of the hidden layer so as to achieve the prediction precision of the model.
2. The NSGA and FELM based cold rolling force prediction method of claim 1, wherein in step 3.3, iterative training of the network is started, and when the number of iterations is 15, the iterative training is ended, and the current FELM model is obtained.
3. The NSGA and FELM based cold rolling force prediction method of claim 1, wherein in step 3.3, iterative training network is started, and when verification set error E is reachedvLess than training set error ErAnd obtaining the current FELM model.
4. The NSGA and FELM based cold rolling force prediction method of claim 1, wherein in step 4.4, whether the algorithm can stop running or not is judged, and when the maximum evolution algebra is set to be 100 and the population evolves to the maximum algebra, the algorithm stops running.
5. The NSGA and FELM based cold rolling force prediction method of claim 1, wherein in step 4.4, whether the algorithm can be stopped is determined, and when the stopping tolerance is set to 0.001, that is, the fitness function value reaches 0.001, the algorithm is stopped.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010963972.6A CN112487700B (en) | 2020-09-15 | 2020-09-15 | Cold rolling force prediction method based on NSGA and FELM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010963972.6A CN112487700B (en) | 2020-09-15 | 2020-09-15 | Cold rolling force prediction method based on NSGA and FELM |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112487700A true CN112487700A (en) | 2021-03-12 |
| CN112487700B CN112487700B (en) | 2022-04-19 |
Family
ID=74920975
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010963972.6A Active CN112487700B (en) | 2020-09-15 | 2020-09-15 | Cold rolling force prediction method based on NSGA and FELM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112487700B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113649420A (en) * | 2021-07-28 | 2021-11-16 | 首钢京唐钢铁联合有限责任公司 | Temper mill rolling force obtaining method and device |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009183969A (en) * | 2008-02-05 | 2009-08-20 | Kobe Steel Ltd | Method for predicting rolling load in cold rolling |
| US20180240018A1 (en) * | 2016-05-19 | 2018-08-23 | Jiangnan University | Improved extreme learning machine method based on artificial bee colony optimization |
| CN108805346A (en) * | 2018-06-04 | 2018-11-13 | 东北大学 | A kind of hot continuous rolling force forecasting method based on more hidden layer extreme learning machines |
| CN109558677A (en) * | 2018-11-29 | 2019-04-02 | 东北大学 | A kind of hot rolling strip crown prediction technique based on data-driven |
| CN110110447A (en) * | 2019-05-09 | 2019-08-09 | 辽宁大学 | It is a kind of to mix the feedback limit learning machine steel strip thickness prediction technique that leapfrogs |
| CN110569605A (en) * | 2019-09-11 | 2019-12-13 | 沈阳农业大学 | Non-glutinous rice leaf nitrogen content inversion model method based on NSGA2-ELM |
| CN111241750A (en) * | 2020-01-14 | 2020-06-05 | 邯郸钢铁集团有限责任公司 | BP network cold-rolled strip steel mechanical property prediction method combining genetic algorithm |
-
2020
- 2020-09-15 CN CN202010963972.6A patent/CN112487700B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2009183969A (en) * | 2008-02-05 | 2009-08-20 | Kobe Steel Ltd | Method for predicting rolling load in cold rolling |
| US20180240018A1 (en) * | 2016-05-19 | 2018-08-23 | Jiangnan University | Improved extreme learning machine method based on artificial bee colony optimization |
| CN108805346A (en) * | 2018-06-04 | 2018-11-13 | 东北大学 | A kind of hot continuous rolling force forecasting method based on more hidden layer extreme learning machines |
| CN109558677A (en) * | 2018-11-29 | 2019-04-02 | 东北大学 | A kind of hot rolling strip crown prediction technique based on data-driven |
| CN110110447A (en) * | 2019-05-09 | 2019-08-09 | 辽宁大学 | It is a kind of to mix the feedback limit learning machine steel strip thickness prediction technique that leapfrogs |
| CN110569605A (en) * | 2019-09-11 | 2019-12-13 | 沈阳农业大学 | Non-glutinous rice leaf nitrogen content inversion model method based on NSGA2-ELM |
| CN111241750A (en) * | 2020-01-14 | 2020-06-05 | 邯郸钢铁集团有限责任公司 | BP network cold-rolled strip steel mechanical property prediction method combining genetic algorithm |
Non-Patent Citations (3)
| Title |
|---|
| YUNLONG WANG等: "Multi-Objective Optimization of Rolling Schedule for Five-Stand Tandem Cold Mill", 《IEEE ACCESS》 * |
| ZIYU HU等: "Multi-parameter deep-perception and many-objective autonomous-control of rolling schedule on high speed cold tandem mill", 《ISA TRANSACTIONS》 * |
| 冀秀梅等: "极限学习机在中厚板轧制力预报中的应用", 《钢铁研究学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113649420A (en) * | 2021-07-28 | 2021-11-16 | 首钢京唐钢铁联合有限责任公司 | Temper mill rolling force obtaining method and device |
| CN113649420B (en) * | 2021-07-28 | 2023-10-24 | 首钢京唐钢铁联合有限责任公司 | Temper mill rolling force obtaining method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112487700B (en) | 2022-04-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109558677A (en) | A kind of hot rolling strip crown prediction technique based on data-driven | |
| CN113094980B (en) | Solder paste printing quality prediction method and system based on IGA-DNN | |
| CN112488315B (en) | Batch scheduling optimization method based on deep reinforcement learning and genetic algorithm | |
| CN112733417B (en) | Abnormal load data detection and correction method and system based on model optimization | |
| CN114386364B (en) | Electronic surface mounting welding quality prediction and process parameter optimization method | |
| CN112364560B (en) | Intelligent prediction method for working hours of mine rock drilling equipment | |
| CN112487700B (en) | Cold rolling force prediction method based on NSGA and FELM | |
| CN109940458B (en) | Method for predicting future wear loss of cutter on line | |
| CN114386659A (en) | Pump pipe optimization method for water system of nuclear power plant | |
| CN114302418A (en) | Method and device for detecting network element equipment performance and computer readable medium | |
| CN113376540B (en) | LSTM battery health state estimation method based on evolutionary attention mechanism | |
| CN114004153A (en) | Penetration depth prediction method based on multi-source data fusion | |
| CN114819178A (en) | Railway construction progress index prediction and online updating method | |
| CN112766548A (en) | Order completion time prediction method based on GASA-BP neural network | |
| CN110705844B (en) | Robust optimization method for job shop scheduling scheme based on non-forced idle time | |
| CN111222627A (en) | Air knife distance data driving prediction method for plating control | |
| CN116484745A (en) | Design method of hot-rolled low-alloy steel for physical metallurgy guidance industry big data mining | |
| CN116542382A (en) | Sewage treatment dissolved oxygen concentration prediction method based on mixed optimization algorithm | |
| CN116611911A (en) | Credit risk prediction method and device based on support vector machine | |
| CN115619028A (en) | Clustering algorithm fusion-based power load accurate prediction method | |
| CN111813669B (en) | Adaptive random test case generation method based on multi-target group intelligence | |
| CN110704956B (en) | Cold rolling mill data-driven technological parameter optimization method | |
| CN113705098A (en) | Air duct heater modeling method based on PCA and GA-BP network | |
| CN112016249A (en) | SCR denitration system bad data identification method based on optimized BP neural network | |
| CN112699229A (en) | Self-adaptive question-pushing method based on deep learning 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |