A method for establishing a knowledge base for continuous casting simulation calculation

Through artificial intelligence algorithm models, a knowledge information of continuous casting experts is collected and processed, and an efficient knowledge base is established, which solves the problems of low efficiency and high cost of traditional methods, and realizes the rapid establishment and accurate maintenance of continuous casting simulation calculation knowledge base.

CN116128057BActive Publication Date: 2024-11-05BAOSTEEL ENG & TECH GRP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202211357679.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-11-01
Publication Date
2024-11-05
Estimated Expiration
2042-11-01

AI Technical Summary

Technical Problem

The prior art is inefficient and costly when establishing a knowledge base for continuous casting simulation, requiring a large amount of manual collection and processing of expert data, and at high maintenance costs.

Method used

Using artificial intelligence algorithm models to collect and refine the knowledge and information of experts in the field of continuous casting, use CRNN+CTC, GMM-HMM, CNN and other models to perform picture and speech recognition, and combine big data and data mining technology to verify and classify information to form Knowledge Base and fault feedback and knowledge updates through online experts.

Benefits of technology

It improves the efficiency and accuracy of the knowledge base establishment, reduces costs, and forms a more comprehensive and extensive knowledge base, solving the problems of slow speed, error-prone and high maintenance costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116128057B_ABST
    Figure CN116128057B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for establishing a knowledge base for continuous casting simulation calculation, comprising the following steps: a model selection step: collecting artificial intelligence algorithm models, and selecting a knowledge collection model from the artificial intelligence model; a knowledge collection step: collecting knowledge information of experts in the field of continuous casting through the knowledge collection model, and converting the knowledge information of experts in the field of continuous casting into text information and storing it in a database; a knowledge purification step: purifying the text information, so that the key knowledge information is obtained after the redundant data in the text information is removed. The present invention uses artificial intelligence to refine the information of continuous casting experts, so that the information of continuous casting experts is stored in the database to form a knowledge base for continuous casting simulation calculation, so that the method is more efficient than traditional input or communication means, thereby reducing the cost of establishing a knowledge base and improving the efficiency of establishment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of continuous casting, and in particular to a method for establishing a knowledge base for continuous casting simulation calculation. Background Art

[0002] Continuous casting is a very important part of the steel industry. When developing core equipment for continuous casting, such as tundishes, crystallizers, and fan-shaped segments, finite element simulation calculations are usually used to optimize the design and research and analysis. However, due to the different levels of experience of the simulation personnel themselves, there are situations where the design cannot be made or the designed simulation scheme is unreasonable. Usually, companies will establish a continuous casting simulation expert knowledge base to assist simulation personnel in designing simulation schemes. The establishment of a continuous casting simulation expert knowledge base needs to be based on the capabilities and knowledge of experts with deep professional knowledge and rich practical experience.

[0003] Chinese patent number CN202110173483.5 discloses a method and device for constructing an expert system knowledge base based on machine learning. The method includes: selecting multiple interrelated fault phenomena to complete the fault description, and generating corresponding knowledge entries in the expert system knowledge base or directly obtaining corresponding maintenance records based on the fault description; obtaining maintenance records generated by reviewing knowledge entries by industry experts; inputting maintenance records into the machine learning model, and updating the fault phenomenon frequency, fault cause frequency and the frequency of occurrence of the above two associations in the expert system knowledge base.

[0004] In the process of establishing the continuous casting knowledge base, it is necessary to manually collect the data of continuous casting experts and record the data in the computer. This method is not only inefficient but also costly. Therefore, it is urgent to design a knowledge base establishment method for continuous casting simulation calculation to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide a method for establishing a knowledge base for continuous casting simulation calculation to solve the above-mentioned deficiencies in the prior art.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] A method for establishing a knowledge base for continuous casting simulation calculation comprises the following steps:

[0008] Model selection step: collect artificial intelligence algorithm models and select knowledge collection models from artificial intelligence models;

[0009] Knowledge collection step: collect the knowledge information of experts in the field of continuous casting through the knowledge collection model, and convert the knowledge information of experts in the field of continuous casting into text information and store it in the database;

[0010] Knowledge purification step: purify the text information to remove redundant data in the text information and obtain key knowledge information;

[0011] Knowledge classification step: extract keywords from key knowledge information, identify semantics, and then classify key knowledge;

[0012] Knowledge verification step: Verify the classified information through big data;

[0013] Simulation application steps: download the verified continuous casting knowledge and apply it to the simulation calculation of continuous casting;

[0014] Simulation fault feedback steps: Faults occurring in the continuous casting simulation calculation are fed back to online experts, who analyze the faults and correct the knowledge in the database.

[0015] Furthermore, in the knowledge collection step, when the model collects picture knowledge information of experts in the field of continuous casting, the CRNN+CTC algorithm is used to extract text from the picture.

[0016] Furthermore, in the knowledge collection step, when the model collects the speech knowledge information of experts in the continuous casting field, the GMM-HMM model is used to recognize the text in the speech and extract the text in the speech.

[0017] Furthermore, in the knowledge collection step, when the model collects journal knowledge information of experts in the continuous casting field, the journals related to the continuous casting field are sorted out from the journal database by searching for journal papers in the related continuous casting field, and then the comprehensive score of the journal is evaluated using data mining technology based on indicators such as the number of citations of the paper and the importance of the publishing journal.

[0018] Furthermore, in the knowledge collection step, the collected text information is used as input, combined with the journal content with high comprehensive scores mined from data, and the CNN algorithm is used to extract reasonable expert knowledge information, ultimately forming a continuous casting simulation expert knowledge base.

[0019] Furthermore, in the knowledge classification step, the keywords in the knowledge information include tundish, crystallizer and sector, and the knowledge information collected in the field of continuous casting includes: material properties, model simplification and boundary condition setting.

[0020] Furthermore, the method for extracting text from an image using the CRNN+CTC algorithm comprises the following steps:

[0021] Identify the text area in the image, identify the boundaries of the text, crop the border outside the text area, and input it into the image;

[0022] Perform text recognition on the input image, search for typos in the text, and then correct the typos based on the semantics of the text;

[0023] The corrected text is stored in the database in the form of segments.

[0024] Furthermore, the method of using the GMM-HMM model to recognize text in speech includes the following steps:

[0025] Input the standard speech into the model, train the standard speech through the model, and form a speech template library;

[0026] The continuous casting expert's speech information is input into the model for preprocessing, and the features in the speech information are extracted;

[0027] The voice features are matched with the template library, and the successfully matched feature information is output as text.

[0028] Further, the image is converted into a convolution feature matrix, and the convolution features are input into the Recurrent Layers. The inputting of the convolution features into the Recurrent Layers includes the following steps: first, the image is scaled to a size of 32×W×3 with a fixed aspect ratio, where W represents an arbitrary width; then, after passing through CNN, it becomes 1×(W / 4)×512; and T=(W / 4) is set for LSTM.

[0029] Furthermore, in the keyword classification of the crystallizer, after the text about the crystallizer is extracted, the collected text information is used as a training set, and the RNN algorithm is used to obtain the reasonable working conditions and other demand information required for the simulation. The obtained reasonable working conditions and other demand information are used to obtain important key information in the reasonable working conditions information through the latent Dirichlet distribution topic model algorithm as input, and then the important key information is combined with the simulation expert knowledge base, and the CNN model algorithm is used to finally generate a simulation plan.

[0030] In the above technical solution, the present invention provides a method for establishing a knowledge base for continuous casting simulation calculation, which has the following beneficial effects:

[0031] (1) The present invention applies artificial intelligence to extract the information of continuous casting experts, so that the information of continuous casting experts is stored in a database to form a knowledge base for continuous casting simulation calculation, making this method more efficient than traditional input or communication methods, thereby reducing the cost of establishing a knowledge base and improving the efficiency of establishment.

[0032] (2) The present invention combines the journal database to search and analyze the literature in the field of related continuous casting knowledge, which can ensure the accuracy of expert knowledge and solve the problem of high subsequent maintenance costs.

[0033] (3) The present invention recognizes the text in the pictures and voice information provided by the continuous casting expert, so that the continuous casting information in the pictures and voice information is converted into text and stored in the database, thereby making the knowledge in the knowledge base more comprehensive and extensive. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings.

[0035] Figure 1 A flow chart is provided for an embodiment of a method for establishing a knowledge base for continuous casting simulation calculation according to the present invention.

[0036] Figure 2 A crystallizer copper plate simulation flow chart is provided for an embodiment of a method for establishing a knowledge base for continuous casting simulation calculation according to the present invention.

[0037] Figure 3 A speech recognition flow chart is provided for an embodiment of a method for establishing a knowledge base for continuous casting simulation calculation according to the present invention. DETAILED DESCRIPTION

[0038] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0039] like Figure 1-3 As shown, a method for establishing a knowledge base for continuous casting simulation calculation provided by an embodiment of the present invention includes the following steps:

[0040] Model selection step: collect artificial intelligence algorithm models and select knowledge collection models from artificial intelligence models;

[0041] Knowledge collection step: collect the knowledge information of experts in the field of continuous casting through the knowledge collection model, and convert the knowledge information of experts in the field of continuous casting into text information and store it in the database;

[0042] Knowledge purification step: purify the text information to remove redundant data in the text information and obtain key knowledge information;

[0043] Knowledge classification step: extract keywords from key knowledge information, identify semantics, and then classify key knowledge;

[0044] Knowledge verification step: Verify the classified information through big data;

[0045] Simulation application steps: download the verified continuous casting knowledge and apply it to the simulation calculation of continuous casting;

[0046] Simulation fault feedback steps: Faults occurring in the continuous casting simulation calculation are fed back to online experts, who analyze the faults and correct the knowledge in the database.

[0047] Specifically, in this embodiment, the following steps are included:

[0048] Model selection step: collect artificial intelligence algorithm models and select knowledge collection models from artificial intelligence models;

[0049] Knowledge collection step: collect the knowledge information of experts in the field of continuous casting through the knowledge collection model, and convert the knowledge information of experts in the field of continuous casting into text information and store it in the database;

[0050] Knowledge purification step: purify the text information to remove redundant data in the text information and obtain key knowledge information;

[0051] Knowledge classification step: extract keywords from key knowledge information, identify semantics, and then classify key knowledge. Keywords in knowledge information include tundish, crystallizer, and fan-shaped segment. Knowledge information collected in the field of continuous casting includes: material properties, model simplification, and boundary condition setting;

[0052] Knowledge verification step: Verify the classified information through big data;

[0053] Simulation application steps: download the verified continuous casting knowledge and apply it to the simulation calculation of continuous casting;

[0054] Simulation fault feedback step: the faults in the simulation calculation of continuous casting are fed back to the online experts, who analyze the faults and correct the knowledge in the database. The existing continuous casting knowledge needs to be continuously induced, continuously de-circulated, and continuously interacted in the process of acquisition and transformation. In order to ensure the credibility of the continuous casting simulation expert knowledge base, a large amount of cost is usually required to maintain it to ensure the correctness, low redundancy and consistency of the knowledge in the knowledge base. The present invention mainly solves the problems of slow speed, easy error and high maintenance cost of the establishment method.

[0055] The present invention provides a method for establishing a knowledge base for continuous casting simulation calculation. The present invention applies artificial intelligence to refine the information of continuous casting experts, so that the information of continuous casting experts is stored in a database to form a knowledge base for continuous casting simulation calculation, so that this method is more efficient than traditional input or communication means, thereby reducing the cost of establishing a knowledge base and improving the efficiency of establishment.

[0056] In another embodiment provided by the present invention, in the knowledge collection step, when the model collects picture knowledge information of experts in the field of continuous casting, the CRNN+CTC algorithm is used to extract text from the picture.

[0057] In another embodiment provided by the present invention, in the knowledge collection step, when the model collects the speech knowledge information of the experts in the continuous casting field, the GMM-HMM model is used to recognize the text in the speech and extract the text in the speech.

[0058] In another embodiment provided by the present invention, in the knowledge collection step, when the model collects journal knowledge information of experts in the continuous casting field, journals related to the continuous casting field are sorted out from the journal database by searching for journal papers in the related continuous casting field, and then the comprehensive score of the journal is evaluated by using data mining technology based on indicators such as the number of citations of the paper and the importance of the journal in which it is published. The collected text information is used as input, combined with the journal content with high comprehensive scores mined by data, and the CNN algorithm is used to extract reasonable expert knowledge information, and finally a continuous casting simulation expert knowledge base is formed.

[0059] In another embodiment provided by the present invention, a method for extracting text from a picture using the CRNN+CTC algorithm comprises the following steps:

[0060] Identify the text area in the image, identify the boundaries of the text, crop the border outside the text area, and input it into the image;

[0061] Perform text recognition on the input image, search for typos in the text, and then correct the typos based on the semantics of the text;

[0062] The corrected text is stored in the database in the form of segments, and the input image size is set to (32, 100, 3). Note that the images mentioned are all in the form of (Height, Width, Channel). The convolutional layer here is a common CNN network, which is used to extract the convolutional feature maps of the input image, that is, convert the image of size (32, 100, 3) into a convolutional feature matrix of size (1, 25, 512), and input the convolutional features into the Recurrent Layers. Inputting the convolutional features into the Recurrent Layers includes the following steps: first, scale the image to 32×W×3 with a fixed aspect ratio, where W represents any width; then, after passing through CNN, it becomes 1×(W / 4)×512; set T=(W / 4) for LSTM, and then input the features into LSTM. Therefore, when processing the input image, it is recommended to scale the height to 32 while maintaining the aspect ratio, so as not to destroy the text details in the image as much as possible. Of course, the input image will also be scaled to a fixed width, but this will definitely cause performance degradation due to the destruction of the shape of the text.

[0063] In another embodiment provided by the present invention, recognizing text in speech using the GMM-HMM model includes the following steps:

[0064] Input the standard speech into the model, train the standard speech through the model, and form a speech template library;

[0065] The continuous casting expert's speech information is input into the model for preprocessing, and the features in the speech information are extracted;

[0066] The voice features are matched with the template library, and the successfully matched feature information is output as text.

[0067] In another embodiment provided by the present invention, during the continuous casting process, the solidification of the ingot in the crystallizer involves complex interface heat transfer, which has a significant impact on the solidification conditions of the ingot and the life of the copper plate. It is necessary to calculate the temperature field of the crystallizer copper plate to provide a theoretical basis for analyzing and solving the quality problems of the ingot in the crystallizer and designing or optimizing the arrangement of thermocouples for temperature measurement for steel leakage prediction. In the process of solving the simulation calculation of the temperature field of the continuous casting crystallizer copper plate, the traditional design institute simulation personnel need to communicate effectively with the on-site personnel or design personnel many times to determine the initial conditions, boundary conditions, crystallizer material physical properties and other information, and combine their own The simulation calculation scheme is designed based on personal experience, but due to the different experience levels of the simulation personnel themselves, there are situations where the design cannot be made or the designed simulation scheme is unreasonable. In the keyword classification of the crystallizer, after extracting the text about the crystallizer, the collected text information is used as the training set, and the RNN algorithm is used to obtain the reasonable working conditions and other demand information required for the simulation. The obtained reasonable working conditions and other demand information are used to obtain the important key information in the reasonable working conditions information through the latent Dirichlet distribution topic model algorithm as input, and then the important key information is combined with the simulation expert knowledge base, and the CNN model algorithm is used to finally generate a simulation scheme. Example

[0068] A method for establishing a knowledge base for continuous casting simulation calculation comprises the following steps: a model selection step: collecting artificial intelligence algorithm models, and selecting a knowledge collection model from the artificial intelligence model; a knowledge collection step: collecting knowledge information of experts in the field of continuous casting through the knowledge collection model, and converting the knowledge information of experts in the field of continuous casting into text information and storing it in a database; a knowledge purification step: purifying the text information, so that the key knowledge information is obtained after the redundant data in the text information is removed; a knowledge classification step: extracting keywords from the key knowledge information, identifying the semantics, and then classifying the key knowledge, wherein the keywords in the knowledge information include tundish, crystallizer and fan-shaped segment, and collecting the knowledge information of experts in the field of continuous casting. The knowledge information includes: material properties, model simplification and boundary condition setting; knowledge verification step: verifying the classified information by means of big data; simulation application step: downloading the verified continuous casting knowledge and applying it to the simulation calculation of continuous casting; simulation fault feedback step: the faults in the simulation calculation of continuous casting are fed back to the online experts, who analyze the faults and correct the knowledge in the database. The existing continuous casting knowledge needs to be continuously induced, continuously de-circulated and continuously interacted in the process of acquisition and transformation. In order to ensure the credibility of the continuous casting simulation expert knowledge base, a large amount of cost is usually required to maintain and ensure the correctness, low redundancy and consistency of the knowledge in the knowledge base. The present invention mainly solves the problems of slow speed, easy error and high maintenance cost of the establishment method. Example

[0069] A method for establishing a knowledge base for continuous casting simulation calculation, this embodiment is further limited on the basis of embodiment 1, wherein, in the knowledge collection step, when the model collects picture knowledge information of experts in the field of continuous casting, the CRNN+CTC algorithm is used to extract the text in the picture; in the knowledge collection step, when the model collects voice knowledge information of experts in the field of continuous casting, the GMM-HMM model is used to recognize the text in the voice and extract the text in the voice; in the knowledge collection step, when the model collects journal knowledge information of experts in the field of continuous casting, from the journal database, by searching for journal papers in the related continuous casting field, the journals on the field of continuous casting are sorted out, and then the data mining technology is used to evaluate the number of citations of the paper, the importance of the published journal and other indicators. The comprehensive score of the journal, taking the collected text information as input, combined with the journal content with high comprehensive score obtained by data mining, using CNN algorithm to extract reasonable expert knowledge information, and finally forming a continuous casting simulation expert knowledge base; the method of extracting text from the picture using CRNN+CTC algorithm includes the following steps: identifying the text area in the picture, identifying the boundary of the text, cutting the border outside the text area, and inputting the picture; performing text recognition on the input picture, searching for typos in the text, and then correcting the typos according to the semantics of the text; storing the corrected text in the database in the form of segments, setting the input image size to (32, 100, 3), and note that the mentioned images are (Height, Width, Channel) In the form of, the convolutional layer here is an ordinary CNN network, which is used to extract the convolutional feature maps of the input image, that is, to convert the image of size (32, 100, 3) into a convolutional feature matrix of size (1, 25, 512), and input the convolutional features into Recurrent Layers. Inputting the convolutional features into Recurrent Layers includes the following steps: first, scale the image to 32×W×3 with a fixed aspect ratio, where W represents any width; then, after passing through CNN, it becomes 1×(W / 4)×512; set T=(W / 4) for LSTM, and then input the features into LSTM.Therefore, when processing the input image, it is recommended to scale the height to 32 while maintaining the aspect ratio, so as not to destroy the text details in the image as much as possible. Of course, the input image will also be scaled to a fixed width, but this will definitely cause performance degradation due to the destruction of the shape of the text; Using the GMM-HMM model to recognize text in speech includes the following steps: inputting standard speech into the model, training the standard speech through the model to form a speech template library; inputting the continuous casting expert voice information into the model for preprocessing, and extracting features from the voice information; matching the voice features with the template library, and outputting the successfully matched feature information as text; In the continuous casting process, the solidification of the ingot in the crystallizer involves complex interface heat transfer, which has an important influence on the solidification conditions of the ingot and the life of the copper plate. It is necessary to calculate the temperature field of the copper plate in the crystallizer to analyze and solve the quality problems of the ingot in the crystallizer, design or optimize the leaky steel The arrangement of the forecast temperature thermocouples provides a theoretical basis; in the process of solving the temperature field simulation calculation of the copper plate of the continuous casting crystallizer, the traditional design institute simulation personnel need to communicate effectively with the on-site personnel or designers many times to determine the initial conditions, boundary conditions, physical properties of the crystallizer materials and other information, and design a simulation calculation plan based on their own experience. However, due to the different experience levels of the simulation personnel themselves, there are cases where the design cannot be made or the designed simulation plan is unreasonable. In the keyword classification of the crystallizer, after extracting the text about the crystallizer, the collected text information is used as a training set, and the RNN algorithm is used to obtain the reasonable working conditions and other demand information required for the simulation. The obtained reasonable working conditions and other demand information are used to obtain the important key information in the reasonable working conditions information through the latent Dirichlet distribution topic model algorithm as input, and then the important key information is combined with the simulation expert knowledge base, and the CNN model algorithm is used to finally generate a simulation plan.

[0070] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for establishing a knowledge base for continuous casting simulation calculation, characterized in that: The following steps are involved: Model selection step: collect artificial intelligence algorithm models and select knowledge collection models from artificial intelligence models; Knowledge collection step: collect the knowledge information of experts in the field of continuous casting through the knowledge collection model, and convert the knowledge information of experts in the field of continuous casting into text information and store it in the database; Knowledge purification step: purify the text information to remove redundant data in the text information and obtain key knowledge information; The knowledge classification step extracts keywords from key knowledge information, identifies semantics, and then classifies key knowledge; In the knowledge classification step, the keywords in the knowledge information include tundish, mold and sector, and the knowledge information collected in the field of continuous casting includes: material properties, model simplification and boundary condition setting; In the keyword classification of the crystallizer, after extracting the text about the crystallizer, the collected text information is used as a training set, and the reasonable working conditions and other demand information required for the simulation are obtained through the RNN algorithm. The obtained reasonable working conditions and other demand information are used to obtain important key information in the reasonable working conditions information through the implicit Dirichlet distribution topic model algorithm as input, and then the important key information is combined with the simulation expert knowledge base, and the CNN model algorithm is used to finally generate a simulation plan; Knowledge verification step: Verify the classified information through big data; Simulation application steps: download the verified continuous casting knowledge and apply it to the simulation calculation of continuous casting; Simulation fault feedback step: The faults in the continuous casting simulation calculation are fed back to the online experts, who analyze the faults and correct the knowledge in the database; In the knowledge collection step, when the model collects the picture knowledge information of the experts in the field of continuous casting, the CRNN+CTC algorithm is used to extract the text in the picture. In the knowledge collection step, when the model collects the speech knowledge information of the experts in the field of continuous casting, the GMM-HMM model is used to recognize the text in the speech and extract the text in the speech. The method for extracting text from a picture using the CRNN+CTC algorithm comprises the following steps: Identify the text area in the image, identify the boundaries of the text, crop the border outside the text area, and input it into the image; Perform text recognition on the input image, search for typos in the text, and then correct the typos based on the semantics of the text; The corrected text is stored in the database in the form of segments; The method of using the GMM-HMM model to recognize text in speech includes the following steps: Input the standard speech into the model, train the standard speech through the model, and form a speech template library; The continuous casting expert's speech information is input into the model for preprocessing, and the features in the speech information are extracted; Match the voice features with the template library, and output the successfully matched feature information in text form; The image is converted into a convolution feature matrix, and the convolution features are input into the Recurrent Layers. The input of the convolution features into the Recurrent Layers includes the following steps: first, the image is scaled to a size of 32×W×3 with a fixed aspect ratio, where W represents an arbitrary width; then, after passing through CNN, it becomes 1×(W / 4)×512; and T=(W / 4) is set for LSTM.

2. A method for establishing a knowledge base for continuous casting simulation calculation according to claim 1, characterized in that: In the knowledge collection step, when the model collects journal knowledge information of experts in the field of continuous casting, the journals related to the field of continuous casting are sorted out from the journal database by searching for journal papers related to the field of continuous casting, and then the comprehensive score of the journal is evaluated by using data mining technology based on indicators such as the number of citations of the papers and the importance of the journals in which they are published; Keywords for searching the journal articles included continuous casting, tundish, mold, and segment.

3. The method for establishing a knowledge base for continuous casting simulation calculation according to claim 1, characterized in that: In the knowledge collection step, the collected text information is used as input, combined with the journal content with high comprehensive scores mined from data, and the CNN algorithm is used to extract reasonable expert knowledge information, ultimately forming a continuous casting simulation expert knowledge base.