CN114626673A - Continuous casting billet quality forecasting method based on Bayesian network - Google Patents
Continuous casting billet quality forecasting method based on Bayesian network Download PDFInfo
- Publication number
- CN114626673A CN114626673A CN202210117535.1A CN202210117535A CN114626673A CN 114626673 A CN114626673 A CN 114626673A CN 202210117535 A CN202210117535 A CN 202210117535A CN 114626673 A CN114626673 A CN 114626673A
- Authority
- CN
- China
- Prior art keywords
- quality
- continuous casting
- network
- bayesian network
- production
- 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
- 238000009749 continuous casting Methods 0.000 title claims abstract description 64
- 238000013277 forecasting method Methods 0.000 title claims abstract description 11
- 238000004519 manufacturing process Methods 0.000 claims abstract description 63
- 238000000034 method Methods 0.000 claims abstract description 56
- 238000011156 evaluation Methods 0.000 claims abstract description 46
- 238000005266 casting Methods 0.000 claims abstract description 40
- 230000007547 defect Effects 0.000 claims abstract description 37
- 230000008569 process Effects 0.000 claims abstract description 28
- 229910000831 Steel Inorganic materials 0.000 claims abstract description 21
- 239000010959 steel Substances 0.000 claims abstract description 21
- 238000004458 analytical method Methods 0.000 claims abstract description 12
- 238000009826 distribution Methods 0.000 claims description 16
- 238000001816 cooling Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 7
- 230000008859 change Effects 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 230000007246 mechanism Effects 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000011160 research Methods 0.000 claims description 5
- 238000007711 solidification Methods 0.000 claims description 5
- 230000008023 solidification Effects 0.000 claims description 5
- 238000003908 quality control method Methods 0.000 claims description 4
- 238000005204 segregation Methods 0.000 claims description 4
- 238000007476 Maximum Likelihood Methods 0.000 claims description 3
- 238000009960 carding Methods 0.000 claims description 3
- 239000000498 cooling water Substances 0.000 claims description 3
- 238000010327 methods by industry Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000005336 cracking Methods 0.000 claims description 2
- 230000005540 biological transmission Effects 0.000 claims 1
- 238000004451 qualitative analysis Methods 0.000 abstract description 6
- 230000001364 causal effect Effects 0.000 abstract description 5
- 238000004445 quantitative analysis Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000009897 systematic effect Effects 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 239000002436 steel type Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B22—CASTING; POWDER METALLURGY
- B22D—CASTING OF METALS; CASTING OF OTHER SUBSTANCES BY THE SAME PROCESSES OR DEVICES
- B22D11/00—Continuous casting of metals, i.e. casting in indefinite lengths
- B22D11/16—Controlling or regulating processes or operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
-
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- 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)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Educational Administration (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Game Theory and Decision Science (AREA)
- Mechanical Engineering (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Algebra (AREA)
- Probability & Statistics with Applications (AREA)
- Computational Mathematics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Manufacturing & Machinery (AREA)
- Primary Health Care (AREA)
- General Factory Administration (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a continuous casting billet quality forecasting method based on a Bayesian network, which relates to the technical field of steel production, and is characterized in that the evaluation indexes of parameters and quality are determined by analyzing the causal relationship between the process parameters and the casting billet quality, and an evaluation system of the production process is constructed; determining the topological structure of the Bayesian network, and realizing qualitative analysis of each node in the continuous casting production process; the parameters of the Bayesian network are learned, so that qualitative analysis of each node in the continuous casting production process is realized; carrying out inference analysis on the Bayesian network by using an inference engine to obtain the evaluation results of production parameters and quality defects; each production parameter (or part of production parameters) is used as an evidence variable to be input into the model, and the quality of the casting blank can be predicted through forward reasoning analysis of a Bayesian network.
Description
Technical Field
The invention relates to the technical field of steel production, in particular to a continuous casting billet quality forecasting method based on a Bayesian network.
Background
The quality of the continuous casting blank is forecasted, core factors influencing quality defects are determined, and important basis can be provided for the quality control of the casting blank. Continuous casting is a process of solidifying molten steel into a casting blank in a steel production flow, and the process comprises the steps of pouring the molten steel into a tundish from a steel ladle, then pouring the molten steel into a crystallizer from the tundish, and then cooling the molten steel through the crystallizer and a secondary cooling zone, wherein a plurality of production parameters exist in each step. In addition, the produced quality defects are various in types, such as center segregation, center porosity, shrinkage cavity, middle crack, center crack, corner crack and the like; meanwhile, the continuous casting blank has an intricate relationship between production parameters and quality defects, and the production parameters and the quality defects are influenced mutually due to non-independent causal relationships. Therefore, how to accurately find the source of the continuous casting billet quality problem and the core index of the production parameter through describing qualitative and quantitative relations among all parameters in the continuous casting process, and accurately and effectively optimize the continuous casting process is a key problem of the casting billet quality prediction.
At present, the continuous casting billet quality forecasting method based on the Bayesian network mainly comprises three methods, namely a mechanism model method, an expert system method, an intelligent method taking data driving as a means and the like. The mechanism model method is to analyze the influence of production parameters on casting blank defects or the cause of quality defects by establishing a solidification heat transfer model, but the accuracy of the mechanism model is lack of a verification means at present; the quality forecasting method based on the expert system does not excavate the quantitative relation between the continuous casting production parameters and the quality defects, and is difficult to find out leading factors and core factors; the intelligent methods represented by data driving are mostly performed by a deep learning method or a mathematical statistics method, and the methods lack the explanation of causal relationship and are not beneficial to the tracing of quality causes.
Disclosure of Invention
Aiming at the technical problems, the invention overcomes the defects of the prior art and provides a continuous casting billet quality forecasting method based on a Bayesian network, which comprises the following steps:
(1) relationship between carding continuous casting production parameters and continuous casting billet quality
Analyzing the influence of each process parameter on the quality defect of the casting blank according to the solidification mechanism and metallurgical knowledge;
(2) establishing a continuous casting process evaluation system
The continuous casting process evaluation system is generally organically composed of selected evaluation indexes and characteristics of all layers of the evaluation indexes, and is divided into a quality defect layer and a production parameter layer;
in consideration of the characteristics of the Bayesian network, performing data preprocessing and normalization on each evaluation index, wherein each evaluation index in the original data is 'continuous' data and needs to be subjected to discretization preprocessing;
(3) casting blank quality prediction model based on Bayesian network
1) Learning the Bayesian network structure of the forecasting model by adopting a structure learning algorithm based on grading search, searching all possible Bayesian network structures, grading each possible structure, and guiding the next round of search according to the grading result until the grading is converged; the scoring function solves the posterior probability of the network structure under a given data set D through a Bayesian formula;
2) learning parameters of a Bayesian network in a casting blank quality forecasting model;
3) logical reasoning of a Bayes network in the casting blank quality forecasting model is carried out by adopting a joint tree reasoning engine;
4) inputting actual production parameters as evidence variables into the model, and realizing the prediction of the casting blank quality through forward reasoning analysis of a Bayesian network;
(4) casting blank quality problem reverse tracing source based on Bayesian network
Similar to the casting blank quality forecasting model, after the structure learning and the parameter learning are completed, the reverse tracing of the complete Bayes network is carried out on the basis of a joint tree inference engine, and the casting blank quality reverse tracing model based on the Bayes network is constructed; and performing reverse traceability reasoning on the Bayes network in the continuous casting production process by taking the quality defect of the casting blank as an evidence variable to obtain the discrete state and the edge probability distribution of each production parameter node, thereby obtaining the reverse traceability evaluation result of the production parameters in the continuous casting process engineering, finding out key parameters influencing the product quality, and providing a basis for product quality control. .
The technical scheme of the invention is further defined as follows:
in the foregoing method for forecasting quality of a continuous casting slab based on a bayesian network, in step (1), each process parameter includes a superheat degree of molten steel, a casting speed, a temperature difference of cooling water of a crystallizer, a cooling parameter of each secondary section, and a steel type, and quality defects of the casting slab mainly include: center segregation, porosity, shrinkage cavity, cracking.
In the foregoing method for forecasting the quality of a continuous casting slab based on a bayesian network, in the step (2), firstly, the components of a steel grade are analyzed according to actual production data; secondly, determining an evaluation index of the quality defect layer according to the quality defect data; and finally, determining the evaluation index of the production parameter layer according to the mutual influence relation among the production parameters by combining the research and analysis of the industrial background.
In the foregoing method for forecasting the quality of a continuous casting slab based on a bayesian network, in the step (2), the discretization pretreatment specifically includes: firstly, carrying out data perspective processing on original data, and determining discretization values of each evaluation index by combining research and analysis of an industrial background; secondly, further analyzing the evaluation indexes based on the result of the data perspective processing and the determined discretization value, and determining the discretization standard of each evaluation index; and finally, normalizing the original data and integrating the normalized data into a normalized data set.
In the foregoing continuous casting slab quality forecasting method based on the bayesian network, in the step (3), the scoring formula is as follows:
In the foregoing continuous casting slab quality forecasting method based on the bayesian network, in step (3), the network structure learning process is as follows:
step 1: determining an initial network structure model;
step 2: scoring each possible network structure change based on a scoring function, and finding out a change with the maximum joint probability of the data set D and the network structure N _ i;
and step 3: modifying the network structure model according to the step 2;
and 4, step 4: and repeating the step 2 and the step 3 until a network structure with the maximum joint probability of the data set D and the network structure is found.
In the foregoing method for forecasting quality of a continuous casting slab based on a bayesian network, in step (3), the main input items of structure learning include: the method comprises the steps of normalizing a discrete data set, the possible discrete state value number of each node, the node sequence and the maximum father node number, wherein the normalized discrete data set is obtained by processing actual continuous casting production data; the possible discrete state value number of each node is determined by an evaluation system; the node sequence is determined according to the sequence of each production parameter and the correlation of quality defects in the continuous casting process, the production parameters are sequenced from the tundish to the crystallizer to the secondary cooling area, and the quality defects are arranged behind all the production parameters.
In the foregoing continuous casting slab quality forecasting method based on a bayesian network, in step (3), on the basis of giving a data set D and obtaining a bayesian network topology, the main steps of parameter learning are as follows:
step 1: determining a prior distribution of network parameters;
and 2, step: based on a given data set D, the posterior distribution of the nodes is calculated according to a Bayesian formula:
Considering that the continuous casting process is complex, the prior distribution of production parameters and quality defects is unknown, and obtaining unbiased estimation of the original distribution by adopting maximum likelihood estimation.
In the foregoing method for forecasting quality of a continuous casting slab based on a bayesian network, in the step (3), the bayesian network is converted into a junction tree in the first step of joint tree inference, the junction tree is a directed tree in nature, the largest complete subtree of the directed tree is converted into each tree node of the junction tree, and the probability of each node is calculated in the second step according to a message passing protocol defined in the junction tree.
The invention has the beneficial effects that: determining the evaluation indexes of the parameters and the quality by analyzing the causal relationship between the process parameters and the casting blank quality, and constructing an evaluation system of the production process; determining the topological structure of the Bayesian network, and realizing qualitative analysis of each node in the continuous casting production process; the parameters of the Bayesian network are learned, so that qualitative analysis of each node in the continuous casting production process is realized; carrying out inference analysis on the Bayesian network by using an inference engine to obtain the evaluation results of production parameters and quality defects; each production parameter (or part of production parameters) is used as an evidence variable to be input into the model, and the quality of the casting blank can be predicted through forward reasoning analysis of a Bayesian network.
Drawings
FIG. 1 is a casting blank quality forecasting process based on a Bayesian network;
fig. 2 is a flow chart of bayesian network structure learning.
Detailed Description
The method for forecasting the quality of the continuous casting billet based on the Bayesian network, as shown in FIG. 1, comprises the following steps:
(1) relationship between carding continuous casting production parameters and continuous casting billet quality
According to the solidification mechanism and metallurgical knowledge, analyzing the influence of various process parameters such as the superheat degree of molten steel, the throwing speed, the temperature difference of cooling water of a crystallizer, cooling parameters of secondary sections, steel grades and the like on the quality defects of casting blanks, wherein the quality defects of the casting blanks mainly comprise: center segregation, porosity, shrinkage cavity and cracks;
(2) establishing continuous casting process evaluation system
When the system is used for selecting indexes and building, the five principles are followed as follows: scientific principle, comprehensive and systematic principle, objective principle, qualitative analysis and quantitative analysis organic combination principle and steel grade difference principle, wherein the scientific principle means that the continuous casting process can be accurately and effectively evaluated; the comprehensive and systematic principle means that the correlation among all indexes and the causal relationship between the indexes and parameters are considered, and the economic benefit of the steel enterprise can be embodied; the objective principle is the aim of comprehensively considering the goals of optimizing the continuous casting process and improving the economic benefit of enterprises for guidance; the organic combination principle of qualitative analysis and quantitative analysis refers to the fact that whether the different evaluation indexes have mutual influence and mutual connection and the mutual influence are researched; the principle of steel grade difference is that when the evaluation indexes of steel grades with similar components are selected, different cooling solidification characteristics of the steel grades are not confused;
the continuous casting process evaluation system is generally organically composed of selected evaluation indexes and characteristics of each layer of the evaluation indexes, and is divided into a quality defect layer and a production parameter layer, and firstly, steel components are analyzed according to actual production data; secondly, determining an evaluation index of the quality defect layer according to the quality defect data; finally, determining the evaluation index of the production parameter layer according to the mutual influence relation among all the production parameters by combining the research and analysis of the industrial background;
considering the characteristics of a Bayesian network, performing data preprocessing and normalization on each evaluation index, wherein each evaluation index in original data is 'continuous' data and needs to be subjected to discretization preprocessing; secondly, further analyzing the evaluation indexes based on the result of the data perspective processing and the determined discretization value, and determining the discretization standard of each evaluation index; finally, normalizing the original data and integrating the normalized data into a normalized data set;
(3) casting blank quality forecasting model based on Bayesian network
1) Learning the Bayesian network structure of the forecasting model by adopting a structure learning algorithm based on score search, searching all possible Bayesian network structures, scoring each possible structure, and guiding the next round of search according to a scoring result until the scoring is converged; the scoring function solves the posterior probability of the network structure under a given data set D through a Bayesian formula, and the scoring formula is as follows:
the network structure learning process is as shown in FIG. 2:
step 1: determining an initial network structure model;
step 2: scoring each possible network structure change based on a scoring function, and finding out a change with the maximum joint probability of the data set D and the network structure N _ i;
and step 3: modifying the network structure model according to the step 2;
and 4, step 4: repeating the step 2 and the step 3 until a network structure with the maximum joint probability of the data set D and the network structure is found;
the main inputs for structure learning include: the method comprises the steps of a normalized discrete data set, the possible discrete state value number of each node, the node sequence and the maximum father node number, wherein the normalized discrete data set is obtained after actual continuous casting production data are processed; the possible discrete state value number of each node is determined by an evaluation system; the node sequence is determined according to the sequence of each production parameter and the correlation of quality defects in the continuous casting process, the production parameters are sequenced from the tundish to the crystallizer and then to the secondary cooling zone, and the quality defects are arranged behind all the production parameters;
2) parameter learning of a Bayesian network in a casting blank quality forecasting model, wherein the parameter learning aims at obtaining edge probability distribution-network parameters in each node of the Bayesian network, realizing quantitative analysis of the Bayesian network and obtaining quantitative expressions of mutual influence and mutual dependency relationship among the nodes, and the parameter learning mainly comprises the following steps on the basis of giving a data set D and obtaining a Bayesian network topological structure:
step 1: determining a prior distribution of network parameters;
step 2: based on a given data set D, the posterior distribution of the nodes is calculated according to a Bayesian formula:
Considering that the continuous casting process is complex, the prior distribution of production parameters and quality defects is unknown, and obtaining unbiased estimation of the original distribution by adopting maximum likelihood estimation;
3) the method comprises the following steps of carrying out logical inference on a Bayes network in a casting blank quality forecasting model, wherein the inference is to inquire the state of each node on the basis of obtaining a complete Bayes network, carry out inference calculation on the probability distribution of each node, output the most possible discrete state of each node, carry out logical inference by adopting a joint tree inference engine, convert the Bayes network into a joint tree in the first step of joint tree inference, the joint tree is a directed tree in nature, convert the largest complete subtree of the directed tree into each tree node of the joint tree, and calculate the probability of each node in the second step according to a defined message transfer protocol in the joint tree;
4) inputting actual production parameters as evidence variables into the model, and realizing the prediction of the casting blank quality through forward reasoning analysis of a Bayesian network;
(4) casting blank quality problem reverse tracing source based on Bayesian network
Similar to the casting blank quality forecasting model, after the structure learning and the parameter learning are completed, the reverse tracing of the complete Bayes network is carried out on the basis of a joint tree inference engine, and the casting blank quality reverse tracing model based on the Bayes network is constructed; and performing reverse traceability reasoning on the Bayes network in the continuous casting production process by taking the quality defect of the casting blank as an evidence variable to obtain the discrete state and the edge probability distribution of each production parameter node, thereby obtaining the reverse traceability evaluation result of the production parameters in the continuous casting process engineering, finding out key parameters influencing the product quality, and providing a basis for product quality control.
In addition to the above embodiments, the present invention may have other embodiments. All technical solutions formed by adopting equivalent substitutions or equivalent transformations fall within the protection scope of the claims of the present invention.
Claims (9)
1. A continuous casting billet quality forecasting method based on a Bayesian network is characterized in that: the method comprises the following steps:
(1) relationship between carding continuous casting production parameters and continuous casting billet quality
Analyzing the influence of each process parameter on the quality defect of the casting blank according to the solidification mechanism and metallurgical knowledge;
(2) establishing continuous casting process evaluation system
The continuous casting process evaluation system is generally organically composed of selected evaluation indexes and characteristics of all layers of the evaluation indexes, and is divided into a quality defect layer and a production parameter layer;
in consideration of the characteristics of the Bayesian network, performing data preprocessing and normalization on each evaluation index, wherein each evaluation index in the original data is 'continuous' data and needs to be subjected to discretization preprocessing;
(3) casting blank quality prediction model based on Bayesian network
1) Learning the Bayesian network structure of the forecasting model by adopting a structure learning algorithm based on grading search, searching all possible Bayesian network structures, grading each possible structure, and guiding the next round of search according to the grading result until the grading is converged; the scoring function solves the posterior probability of the network structure under a given data set D through a Bayesian formula;
2) learning parameters of a Bayes network in a casting blank quality forecasting model;
3) logical reasoning of a Bayes network in the casting blank quality forecasting model is carried out by adopting a joint tree reasoning engine;
4) inputting actual production parameters as evidence variables into the model, and realizing the prediction of the casting blank quality through forward reasoning analysis of a Bayesian network;
(4) casting blank quality problem reverse tracing source based on Bayes network
Similar to the casting blank quality forecasting model, after the structure learning and the parameter learning are completed, the reverse tracing of the complete Bayes network is carried out on the basis of a joint tree inference engine, and the casting blank quality reverse tracing model based on the Bayes network is constructed; and performing reverse traceability reasoning on the Bayes network in the continuous casting production process by taking the quality defect of the casting blank as an evidence variable to obtain the discrete state and the edge probability distribution of each production parameter node, thereby obtaining the reverse traceability evaluation result of the production parameters in the continuous casting process engineering, finding out key parameters influencing the product quality, and providing a basis for product quality control.
2. The method for forecasting the quality of the continuous casting billet based on the Bayesian network as claimed in claim 1, wherein: in the step (1), the technological parameters comprise molten steel superheat degree, throwing speed, crystallizer cooling water temperature difference, secondary cooling parameters of all sections and steel grades, and the quality defects of the casting blank mainly comprise: center segregation, porosity, shrinkage cavity, cracking.
3. The method for forecasting the quality of the continuous casting billet based on the Bayesian network as claimed in claim 1, wherein: in the step (2), firstly, the components of the steel grade are analyzed according to actual production data; secondly, determining an evaluation index of the quality defect layer according to the quality defect data; and finally, determining the evaluation index of the production parameter layer according to the mutual influence relation among the production parameters by combining the research and analysis of the industrial background.
4. The method for forecasting the quality of the continuous casting billet based on the Bayesian network as claimed in claim 1, wherein: in the step (2), the discretization pretreatment specifically comprises: firstly, carrying out data perspective processing on original data, and determining discretization values of each evaluation index by combining research and analysis of an industrial background; secondly, further analyzing the evaluation indexes based on the result of the data perspective processing and the determined discretization value, and determining the discretization standard of each evaluation index; and finally, normalizing the original data and integrating the normalized data into a normalized data set.
5. The method for forecasting the quality of the continuous casting billet based on the Bayesian network as claimed in claim 1, wherein: in the step (3), the scoring formula is as follows:
6. The method for forecasting the quality of the continuous casting billet based on the Bayesian network as claimed in claim 1, wherein: in the step (3), the network structure learning process is as follows:
step 1: determining an initial network structure model;
and 2, step: scoring each possible network structure change based on a scoring function, and finding out a change with the maximum joint probability of the data set D and the network structure N _ i;
and step 3: modifying the network structure model according to the step 2;
and 4, step 4: and repeating the step 2 and the step 3 until a network structure with the maximum joint probability of the data set D and the network structure is found.
7. The method for forecasting the quality of the continuous casting billet based on the Bayesian network as claimed in claim 1, wherein: in the step (3), the main input items of the structure learning include: the method comprises the steps of normalizing a discrete data set, the possible discrete state value number of each node, the node sequence and the maximum father node number, wherein the normalized discrete data set is obtained by processing actual continuous casting production data; the possible discrete state value number of each node is determined by an evaluation system; the node sequence is determined according to the sequence of each production parameter and the correlation of quality defects in the continuous casting process, the production parameters are sequenced from the tundish to the crystallizer and then to the secondary cooling zone, and the quality defects are arranged behind all the production parameters.
8. The method for forecasting the quality of the continuous casting billet based on the Bayesian network as claimed in claim 1, wherein: in the step (3), on the basis of giving the data set D and obtaining the bayesian network topology, the main steps of parameter learning are as follows:
step 1: determining a prior distribution of network parameters;
and 2, step: based on a given data set D, the posterior distribution of the nodes is calculated according to a Bayesian formula:
Considering that the continuous casting process is complex, the prior distribution of production parameters and quality defects is unknown, and obtaining unbiased estimation of the original distribution by adopting maximum likelihood estimation.
9. The method for forecasting the quality of the continuous casting billet based on the Bayesian network as claimed in claim 1, wherein: in the step (3), the first step of the joint tree reasoning is to convert the Bayesian network into a junction tree, the junction tree is a undirected tree in nature, the largest complete subtree of the undirected tree is converted into each tree node of the junction tree, and the second step is to calculate the probability of each node according to the defined message transmission protocol in the junction tree.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210117535.1A CN114626673A (en) | 2022-02-08 | 2022-02-08 | Continuous casting billet quality forecasting method based on Bayesian network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210117535.1A CN114626673A (en) | 2022-02-08 | 2022-02-08 | Continuous casting billet quality forecasting method based on Bayesian network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114626673A true CN114626673A (en) | 2022-06-14 |
Family
ID=81898771
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210117535.1A Pending CN114626673A (en) | 2022-02-08 | 2022-02-08 | Continuous casting billet quality forecasting method based on Bayesian network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114626673A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115345434A (en) * | 2022-07-15 | 2022-11-15 | 华院计算技术(上海)股份有限公司 | Improved dynamic data mining method and device for continuous casting quality judgment model |
| CN116579650A (en) * | 2023-04-28 | 2023-08-11 | 华院计算技术(上海)股份有限公司 | Continuous casting quality judging method and device, computer readable storage medium and terminal |
| CN117689256A (en) * | 2023-12-12 | 2024-03-12 | 中南大学 | Aluminum alloy casting product quality tracing method |
-
2022
- 2022-02-08 CN CN202210117535.1A patent/CN114626673A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115345434A (en) * | 2022-07-15 | 2022-11-15 | 华院计算技术(上海)股份有限公司 | Improved dynamic data mining method and device for continuous casting quality judgment model |
| CN116579650A (en) * | 2023-04-28 | 2023-08-11 | 华院计算技术(上海)股份有限公司 | Continuous casting quality judging method and device, computer readable storage medium and terminal |
| CN116579650B (en) * | 2023-04-28 | 2024-04-26 | 华院计算技术(上海)股份有限公司 | Continuous casting quality judging method and device, computer readable storage medium and terminal |
| CN117689256A (en) * | 2023-12-12 | 2024-03-12 | 中南大学 | Aluminum alloy casting product quality tracing method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114626673A (en) | Continuous casting billet quality forecasting method based on Bayesian network | |
| Cemernek et al. | Machine learning in continuous casting of steel: A state-of-the-art survey | |
| Wu et al. | Multi-objective optimisation of machining process parameters using deep learning-based data-driven genetic algorithm and TOPSIS | |
| CN113420157A (en) | Steel product surface longitudinal crack defect traceability analysis method based on knowledge graph | |
| CN111488946A (en) | Radar servo system fault diagnosis method based on information fusion | |
| CN111651931A (en) | Blast furnace fault diagnosis rule derivation method based on deep neural network | |
| CN113469241A (en) | Product quality control method based on process network model and machine learning algorithm | |
| Deng et al. | A data-drivenparameter planning method for structural Parts NC machining | |
| CN110245390B (en) | Automobile engine oil consumption prediction method based on RS-BP neural network | |
| CN112418522B (en) | Industrial heating furnace steel temperature prediction method based on three-branch integrated prediction model | |
| Sugisawa et al. | Machining sequence learning via inverse reinforcement learning | |
| Acosta et al. | Machine learning in intelligent manufacturing system for optimization of production costs and overall effectiveness of equipment in fabrication models | |
| CN114462566B (en) | Method for realizing real-time determination of optimal decision action by intelligent real-time decision system | |
| JP3875875B2 (en) | Operation analysis apparatus, method, computer program, and computer-readable storage medium in manufacturing process | |
| CN115174263B (en) | Attack path dynamic decision method and device | |
| CN115456264B (en) | Endpoint carbon content and endpoint temperature prediction method for small and medium-sized converter | |
| Surekha et al. | Design and development of knowledge base scheme for chromite-based resin bonded sand core system | |
| Cateni et al. | The importance of variable selection for neural networks-based classification in an industrial context | |
| Nieves et al. | Combination of machine-learning algorithms for fault prediction in high-precision foundries | |
| JP2018147103A (en) | Model learning device, controlled variable calculation device, and program | |
| CN114331139A (en) | Campus risk monitoring and early warning system based on block chain network | |
| Chen et al. | The use of improved algorithm of adaptive neuro-fuzzy inference system in optimization of machining parameters | |
| CN112183642A (en) | Method and system for detecting coal consumption of cement firing based on random forest model | |
| CN116933449B (en) | Transmission gear robust design optimization method based on gradient auxiliary learning function | |
| Baraka et al. | Perpetual learning framework based on type-2 fuzzy logic system for a complex manufacturing process |
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 |