Intelligent configuration method and system for steelmaking continuous casting digital twinning application

Abstract

The invention provides a steelmaking continuous casting digital twinning application intelligent configuration method and system, and the method comprises the steps: building a function logic through adjusting nodes and a node group according to a production plan, configuring the function parameters of the function logic, and recording the use information of the nodes; adding the function parameters and the use information into a data set, and training a BERT model through the data set; judging whether the prediction of the BERT model is accurate or not according to the information of the data set, and if so, executing the next step; if not, the last step is executed again; judging whether the function is realized; if the result is yes, the current log record and the function parameters are recorded in the corresponding json file, configuration is completed, and function logic is adjusted according to the configuration; if not, the first step is executed again. According to the method, verified functions and code logic are used, so that the possibility of bug occurrence is reduced, and efficient code assets which can be quickly retrieved and reused can be continuously accumulated.

Description

Technical Field

[0001] The present invention belongs to the technical field of digital twin development, and specifically, relates to an intelligent configuration method and system for digital twin applications in steelmaking and continuous casting. Background Art

[0002] The construction of digital twin applications, combined with artificial intelligence, can greatly improve industrial production efficiency and achieve cost reduction and efficiency improvement. At a time when all companies are focusing on the construction of digital twin applications, how to quickly develop, quickly iterate, and enhance system reusability has become an urgent problem to be solved.

[0003] At present, the most common coding solution on the market to improve efficiency is to encapsulate the code into nodes, make it visual, modular, and interactively configured centrally, and ultimately achieve the goal of reducing learning costs and speeding up iterations. However, in factories of different industries, the process flows are very different, and general solutions have too much redundancy and are difficult to combine with specific industries; and there is a lack of intelligent processing during the development process, and the cost of retrieval and learning is still high, resulting in limited efficiency improvements and difficulty in logic reuse. Therefore, how to reduce learning and retrieval costs and increase development efficiency has become one of the research focuses in this field.

[0004] In other words, although the current digital twin application configuration methods have used node-based visual programming to reduce the learning cost of developers, the node-based visual programming method has high requirements for retrieval; moreover, the existing solutions lack specificity for the development of digital twin applications for steelmaking and continuous casting scenarios, have too much redundant content that affects development efficiency, and lack modular presentation of nodes and commonly used functions; when designing functions, the speed of locating specific nodes is slow.

[0005] Patent document CN117667055A provides a low-code digital twin scene interaction configuration method and system. The solution builds scenes based on the digital twin low-code platform, configures the template type of scene interaction through the interactive template configuration function, and after selecting the template type, uses a visual drag-and-drop method to sequentially configure the node type and scene interaction, so as to achieve the construction of a complete node-interactive digital twin scene using a low-code method. The method / system can achieve a high degree of integration of the interactive configuration in the digital twin scene by unifying the interactive configurations scattered in the digital twin scene in the interactive rule function, thereby reducing the complexity of the scene configuration and improving the efficiency of digital twin scene arrangement;

[0006] Its disadvantages are that it lacks specificity in the development of digital twin functions for steelmaking and continuous casting scenarios, and there is too much redundant content in the configuration; although the complexity of scene configuration is reduced from the perspective of configuration integration, the node interactive programming method still has a high learning cost, and there are problems such as low node retrieval efficiency that need to be solved.

[0007] Patent document CN117075884A discloses a digital processing system and method based on visual scripts. The system provided by the solution includes a visual script editing module, a view analysis module, a node configuration module and a digital processing module; the output end of the visual script editing module is connected to the input end of the view analysis module; the output end of the view analysis module is connected to the input end of the node configuration module; the output end of the node configuration module is connected to the input end of the digital processing module. It can realize visual programming, write logic through a graphical interface, and use nodes and lines to represent logical relationships, making the logical structure clearer and easier to maintain, saving developers' time and energy;

[0008] Its disadvantages are that it does not show the configuration for steelmaking and continuous casting scenarios, lacks intelligent processing operations, and has a high learning cost. Summary of the invention

[0009] In view of the defects in the prior art, the object of the present invention is to provide an intelligent configuration method and system for digital twin applications in steelmaking and continuous casting.

[0010] According to the present invention, a steelmaking and continuous casting digital twin application intelligent configuration method is provided, comprising:

[0011] Step S1: According to the production plan, construct the functional logic by adjusting the nodes and node groups, and then configure the functional parameters of the functional logic, and record the usage information of the nodes;

[0012] Step S2: adding the function parameters and usage information into a data set, and training a BERT model using the data set;

[0013] Step S3: judging whether the prediction of the BERT model is accurate according to the information of the data set, if the result is yes, executing step S4; if the result is no, re-executing step S2;

[0014] Step S4: Determine whether the function is implemented; if the result is yes, record the current log records and function parameters in the corresponding json file, complete the configuration and adjust the function logic according to the configuration; if the result is no, re-execute step S1.

[0015] Preferably, in step S1, the function parameters include: function name, function usage scenario and function label;

[0016] The usage information includes the nodes used in the function implementation process, the number of node groups, function labels, node group labels and node connection methods.

[0017] Preferably, the data set is the SCCTDS data set; the SCCTDS data set is used to train the BERT model;

[0018] In step S2, the training of the BERT model includes:

[0019] Step A1: Add data to the SCCTDS dataset; the data includes: sentence and tag;

[0020] The sentences include: "I need to plan and schedule all equipment", "I have a production plan that needs to be simulated", "Set 5 different perspectives facing the converter", and "I need the continuous caster to simulate the billet drawing speed and billet cutting process";

[0021] The tags include: "Planning and Scheduling", "Perspective Switching" and "Production Operation Simulation";

[0022] Step A2: word segmentation processing of the data to obtain training parameters;

[0023] Step A3: calling the Trainer class to train the BERT model through training parameters;

[0024] In the step A2, the text data in the data is marked using the BERT marker bert-base-uncased, thereby generating a parameter input_ids and a parameter attention_masks; the parameter input_ids and the parameter attention_masks are converted into PyTorch tensors, thereby synthesizing a PyTorch data frame;

[0025] In step A3, based on the PyTorch data frame, the Trainer class is called to train the BERT model through training parameters.

[0026] Preferably, in the json file of step S4, the json format of the node includes: "name", "type", "connections", "usage_count", "scene" and "tags" fields;

[0027] The "name" is the name of the storage node;

[0028] The "type" attribute of the node json is "node", indicating that the type of the object is a node; the "connections" records the node names that are allowed to be connected upstream and downstream of the node, including two arrays "upstream" and "downstream";

[0029] The "upstream" array contains the names of the upstream connectable nodes. Specifically, if there is no connectable upstream node, "upstream" displays null; the "downstream" array contains the names of the downstream connectable nodes. If there is no connectable downstream node, "downstream" displays null;

[0030] The "usage_count" stores an integer indicating the number of times it is used;

[0031] The "scene" is an array for listing the scenes using the node, including: "steelmaking scene", "converter scene", and "continuous casting scene";

[0032] The "tags" also stores an array for recording tags related to the node, including: "planning and scheduling", "steel ladle visibility";

[0033] The node group json format includes: "name", "type", "nodeconnections", "usage_count", "scene" and "tags" fields;

[0034] "name" stores the name of the node group. The "type" attribute of the node group json is "nodegroup", indicating that the type of the object is a node group; "nodeconnections" is an object used to record each node used by the node group and record the connection method of the node in the form of an array; "usage_count" stores an integer indicating the number of times it is used; "scene" is an array used to list the scenarios in which the node group is used, including: "steelmaking scenario", "converter scenario", and "continuous casting scenario"; "tags" stores an array used to record tags related to the node, including: "planning and scheduling", "steel ladle visibility";

[0035] The function json format contains the fields "name", "type", "nodeconnections", "usage_count", "scene" and "tags". "name" stores the name of the function. The "type" attribute of the function json is "founction", indicating that the type of the object is a function. "nodeconnections" is an object used to record the nodes and node groups used by the function, and to record the connection methods of the nodes and node groups in the form of an array. "usage_count" stores an integer indicating the number of times the function is used. "scene" is an array used to list the scenarios in which the function is used, including "steelmaking scenario", "converter scenario" and "continuous casting scenario". "tags" stores an array used to record tags related to the function, including "planning and scheduling" and "steel ladle visibility".

[0036] In step S4, based on the input tag, the nodes and node connection methods related to the input tag function are displayed to form recommendation information one; based on the usage information, the usage frequency of the nodes is sorted to form recommendation information two; the recommendation information one and the recommendation information two are combined to obtain the total recommendation information; through the total recommendation information, the configuration is completed and the function logic is adjusted according to the configuration.

[0037] According to the present invention, a steelmaking and continuous casting digital twin application intelligent configuration system is provided, comprising:

[0038] Module M1: According to the production plan, the functional logic is constructed by adjusting the nodes and node groups, and then the functional parameters of the functional logic are configured, and the usage information of the nodes is recorded;

[0039] Module M2: adding the function parameters and usage information to the data set, and training the BERT model through the data set;

[0040] Module M3: Based on the information of the data set, determine whether the prediction of the BERT model is accurate. If the result is yes, execute module M4; if the result is no, re-trigger module M2;

[0041] Module M4: Determine whether the function is implemented; if the result is yes, the current log records and function parameters are recorded in the corresponding json file, the configuration is completed and the function logic is adjusted according to the configuration; if the result is no, module M1 is re-triggered.

[0042] Preferably, in the module M1, the function parameters include: function name, function usage scenario and function label;

[0043] The usage information includes the nodes used in the function implementation process, the number of node groups, function labels, node group labels and node connection methods.

[0044] Preferably, the data set is the SCCTDS data set; the SCCTDS data set is used to train the BERT model;

[0045] In the module M2, the training of the BERT model includes:

[0046] Module A1: adding data to the SCCTDS dataset; the data includes: sentence and tag;

[0047] The sentences include: "I need to plan and schedule all equipment", "I have a production plan that needs to be simulated", "Set 5 different perspectives facing the converter", and "I need the continuous caster to simulate the billet drawing speed and billet cutting process";

[0048] The tags include: "Planning and Scheduling", "Perspective Switching" and "Production Operation Simulation";

[0049] Module A2: word segmentation processing of the data to obtain training parameters;

[0050] Module A3: Call the Trainer class to train the BERT model through training parameters;

[0051] In the module A2, the BERT tagger bert-base-uncased is used to tag the text data in the data, thereby generating a parameter input_ids and a parameter attention_masks; the parameter input_ids and the parameter attention_masks are converted into PyTorch tensors, thereby synthesizing a PyTorch data frame;

[0052] In the module A3, based on the PyTorch data frame, the Trainer class is called to train the BERT model through training parameters.

[0053] Preferably, in the module M4, the json format of the node includes: "name", "type", "connections", "usage_count", "scene" and "tags" fields;

[0054] The "name" is the name of the storage node;

[0055] The "type" attribute of the node json is "node", indicating that the type of the object is a node; the "connections" records the node names that are allowed to be connected upstream and downstream of the node, including two arrays "upstream" and "downstream";

[0056] The "upstream" array contains the names of the upstream connectable nodes. Specifically, if there is no connectable upstream node, "upstream" displays null; the "downstream" array contains the names of the downstream connectable nodes. If there is no connectable downstream node, "downstream" displays null;

[0057] The "usage_count" stores an integer indicating the number of times it is used;

[0058] The "scene" is an array for listing the scenes using the node, including: "steelmaking scene", "converter scene", and "continuous casting scene";

[0059] The "tags" also stores an array for recording tags related to the node, including: "planning and scheduling", "steel ladle visibility";

[0060] The node group json format includes: "name", "type", "nodeconnections", "usage_count", "scene" and "tags" fields;

[0061] "name" stores the name of the node group. The "type" attribute of the node group json is "nodegroup", indicating that the type of the object is a node group; "nodeconnections" is an object used to record each node used by the node group and record the connection method of the node in the form of an array; "usage_count" stores an integer indicating the number of times it is used; "scene" is an array used to list the scenarios in which the node group is used, including: "steelmaking scenario", "converter scenario", and "continuous casting scenario"; "tags" stores an array used to record tags related to the node, including: "planning and scheduling", "steel ladle visibility";

[0062] The function json format contains the fields "name", "type", "nodeconnections", "usage_count", "scene" and "tags". "name" stores the name of the function. The "type" attribute of the function json is "founction", indicating that the type of the object is a function. "nodeconnections" is an object used to record the nodes and node groups used by the function, and to record the connection methods of the nodes and node groups in the form of an array. "usage_count" stores an integer indicating the number of times the function is used. "scene" is an array used to list the scenarios in which the function is used, including "steelmaking scenario", "converter scenario" and "continuous casting scenario". "tags" stores an array used to record tags related to the function, including "planning and scheduling" and "steel ladle visibility".

[0063] In the module M4, according to the input tag, the nodes and node connection methods related to the input tag function are displayed to form recommendation information one; according to the usage information, the usage frequency of the nodes is sorted to form recommendation information two; the recommendation information one and the recommendation information two are combined to obtain the total recommendation information; through the total recommendation information, the configuration is completed and the functional logic is adjusted according to the configuration.

[0064] According to a computer-readable storage medium storing a computer program provided by the present invention, it is characterized in that when the computer program is executed by a processor, the steps of the intelligent configuration method of the digital twin application of steelmaking and continuous casting are implemented.

[0065] An electronic device provided according to the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The electronic device is characterized in that when the computer program is executed by the processor, the steps of the intelligent configuration method of the digital twin application of steelmaking and continuous casting are implemented.

[0066] Compared with the prior art, the present invention has the following beneficial effects:

[0067] 1. The system provided by the present invention integrates multiple systems such as enterprise resource planning, i.e. ERP, manufacturing execution system, i.e. MES, and quality management system, i.e. QMS, making it easier for developers to retrieve required data; the system includes multiple equipment modules and preset functions, enabling developers to build steelmaking and continuous casting digital twin scenarios and develop required functions more quickly.

[0068] 2. The present invention improves the reusability of relevant code logic in the process of steelmaking and continuous casting digital twin application development, saving a lot of time consumed by repeated engineering modifications.

[0069] 3. The present invention uses verified functions and code logic, reduces the possibility of bugs, and can continuously accumulate efficient, quickly retrievable and reusable code assets.

[0070] 4. The json file storage format provided by the present invention is the basis for implementing the intelligent configuration method of the digital twin application of steelmaking and continuous casting. Using this storage method, the classification of nodes, node groups and functions can be more accurate, the association can be clearer, and it can be easier to match with the labels summarized by natural semantic recognition, thereby achieving more targeted and effective screening.

[0071] 5. The screening method provided by the present invention automatically labels nodes, node groups or specific functions through a user behavior detection method, uses the BERT algorithm in AI recognition to extract labels from the functional requirement text, matches and screens the labels recognized by AI with the labels of nodes, node groups or functions to achieve intelligent detection, and then the developer selects the labels and configures the parameters to reduce the retrieval cost of the developer.

[0072] 6. The present invention adopts natural language parsing and developer behavior detection technology to classify and sort the nodes and optimize the function configuration process. This method can achieve natural language demand recognition, one-click configuration of existing steelmaking and continuous casting digital twin functions, sorting of node usage frequency, screening of commonly used nodes or combinations, and screening of specific function node groups.

[0073] 7. The present invention realizes the maximum reuse of developed functions, fully reduces the difficulty of developing retrieval nodes, reduces the experimental cost during node selection, and greatly improves development efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0075] Figure 1 A schematic diagram of a standard flow chart for using the functional configuration module provided by the present invention;

[0076] Figure 2 A schematic diagram of the configuration process of an embodiment provided by the present invention;

[0077] Figure 3 A schematic diagram of the relationship between nodes, node groups and functions provided by the present invention;

[0078] Figure 4 The present invention provides an embodiment of the use process schematic diagram. DETAILED DESCRIPTION

[0079] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0080] The present invention provides an intelligent configuration optimization method in the process of steelmaking and continuous casting digital twin application development, adopts natural language processing and developer behavior detection technology, and automatically matches the most suitable function and related node group for the steelmaking and continuous casting scenario after inputting the functional requirements, thereby improving the development efficiency of developers in realizing the digital twin function of the steelmaking and continuous casting scenario and reducing the learning cost. Specifically, the present invention extracts keywords through the natural language processing AI algorithm, and calculates the node usage frequency based on the developer behavior detection to realize the matching selection of the required function and the nodes associated with its development process. Based on the steelmaking and continuous casting scenario, the basic functions are configured with one click after the scenario is imported; for other functions, the function-related nodes are selected through the AI ​​algorithm to obtain the relevant logic modules and node groups for the realization of the required function, thereby reducing the difficulty of retrieval.

[0081] Embodiment 1: According to a steelmaking continuous casting digital twin application intelligent configuration method provided by the present invention, the setting process of the steelmaking continuous casting production planning and scheduling function is as follows: Figure 2 As shown, including:

[0082] Step 1: Create an empty function template, call nodes and node groups to implement the required function logic; that is, according to the production plan, build the function logic by adjusting nodes and node groups, and then configure the function parameters of the function logic and record the usage information of the nodes;

[0083] In order to realize the production planning and scheduling function, it is first necessary to build multiple node groups through nodes, such as the crane position drive node group, the crane lifting crane drive node group, the trolley position drive node group, the converter opening and closing drive node group, etc.

[0084] Taking the construction of the vehicle position driving node group as an example, it is necessary to use related nodes, such as obtaining relative transformation, setting relative change, moving the vehicle to, branch judgment and other nodes, to build the following logic: read the data in the interface every second to get the actual position of the vehicle. If the position of the vehicle in the three-dimensional scene is inconsistent with the actual position of the vehicle, move the vehicle to the corresponding position. If lifting or lowering is read, execute the corresponding vehicle animation.

[0085] By building the crane position driven node group, after building the remaining node groups, combine them together according to conditions, such as when the crane is lowered, the converter is opened. Finally, the production planning and scheduling function is formed. Among them, when building the logic, you can set the function operation related parameters, such as the crane style, trolley style, converter rotation angle, converter rotation time and other configurable parameters. After the logic is built, the developer can still modify the parameter value and make detailed adjustments.

[0086] In other words, calling nodes and node groups realizes the expected production planning and scheduling functions. The functional template is the core framework for realizing the production planning and scheduling logic by calling nodes and node groups. It has a flexible structure and scalability and can adapt to different production needs and constraints;

[0087] In the functional template, in order to realize the production planning and scheduling logic, it is necessary to define specific attributes such as input, output, and interface; then, by defining node groups, multiple related nodes are organized into a higher-level process unit to support complex execution logic such as sequential, parallel or conditional execution. The functional template can not only flexibly adapt to the needs of various production scenarios, but also use API interfaces to integrate with external systems to ensure the logical implementation of production planning and scheduling.

[0088] Step 2: Configure or set the function parameters;

[0089] The function parameters include: function name, function usage scenario, function label and function operation related parameters.

[0090] The developer behavior detection method is a method for automatically tracking, recording and analyzing developer behavior through systematic means. When developers use nodes, node groups and functions, each operation is automatically captured through embedded event listeners, including clicks, node connections, function calls, node group configurations and other behaviors, and these behavior data are recorded in logs. The collected data is cleaned, analyzed and classified, and stored in json files of nodes, node groups and functions in step 6. That is, the user behavior detection system records the number of nodes, node groups, function labels, node group labels, and node connection methods used in the process of implementing the function;

[0091] Step 3: Add the label, i.e. the function label in the function parameter, to the SCCTDS dataset and retrain the BERT model;

[0092] Specifically, the SCCTDS dataset, whose full name is "Steelmaking Continuous Casting TagDataSet", contains text corpus and corresponding tags of business requirements related to steelmaking continuous casting and general requirements of digital twin virtual scenes. This method uses the BERT model trained by the SCCTDS dataset to realize the function of extracting tags from the requirement text;

[0093] Specifically, training the BERT model is divided into the following steps:

[0094] The first step is to add new data to the existing SCCTDS dataset to obtain a new dataset;

[0095] The data includes: sentence, tag;

[0096] The sentences include: "I need to plan and schedule all equipment", "I have a production plan that needs to be simulated".

[0097] The tags include: "planning and scheduling";

[0098] The second step is to use the BertTokenizer method to perform word segmentation on the data content of the new data set, and use pandas to process the data frame in the subsequent processing;

[0099] Use the pre-trained BERT tagger bert-base-uncased to tokenize the text data, generate two new parameters, input_ids and attention_masks, convert them into PyTorch tensors, and merge them into a data frame;

[0100] The third step is to reload the pre-trained BERT model bert-base-uncased and refresh the previously used classification linear layer. Modify the number of tags in the num_tags parameter of the BERT model, use the PyTorch dataset class CustomDataset, let PyTorch's DataLoader read the data, and finally modify the training parameters in the TrainingArguments class and call the Trainer class to retrain the BERT model.

[0101] Step 4: For the newly added function tag "Planning and Scheduling", enter the relevant demand description to check whether the prediction of the BERT model is accurate; that is, determine whether the prediction of the BERT model is accurate. If the result is yes, execute step 5; if the result is no, re-execute step 3;

[0102] Step 5: After the prediction is completed, that is, after the prediction is accurate, test whether the change of the corresponding parameters can cause the change of function implementation; in other words, filter out the "Planning and Scheduling" tag, set the corresponding parameters, and check whether there is a corresponding change;

[0103] Step 6: Run the function to check whether the data can be successfully accessed and whether the operation mode is consistent with the relevant data. If consistent, it means that the function has been debugged and the records can be confirmed. The log records in the developer behavior detection and the nodes, node groups and function information in the manual records are recorded in the corresponding json file. If inconsistent, re-execute step 1;

[0104] Specifically, in order to realize intelligent configuration, the present invention creates a json file storage format to express the relationship between nodes, node groups and functions, such as Figure 3 To make the related information clear and easy to check, the sample data storage format is as follows:

[0105] The JSON format of a node includes the following fields: "name", "type", "connections", "usage_count", "scene" and "tags".

[0106] The "name" is the name of the storage node;

[0107] The "type" attribute of the node json is "node", indicating that the type of the object is a node; the "connections" records the node names that are allowed to be connected upstream and downstream of the node, including two arrays "upstream" and "downstream";

[0108] The "upstream" array contains the names of the upstream connectable nodes. Specifically, if there is no connectable upstream node, "upstream" displays null; the "downstream" array contains the names of the downstream connectable nodes. If there is no connectable downstream node, "downstream" displays null;

[0109] The "usage_count" stores an integer indicating the number of times it is used;

[0110] The "scene" is an array for listing the scenes where the node is used, including "steelmaking scene", "converter scene", "continuous casting scene" and other scenes;

[0111] The "tags" also stores an array for recording tags related to the node, including multiple tags such as "plan scheduling" and "steel ladle visibility";

[0112] The node group json format includes: "name", "type", "nodeconnections", "usage_count", "scene" and "tags" fields;

[0113] "name" stores the name of the node group. The "type" attribute of the node group json is "nodegroup", indicating that the type of the object is a node group. "nodeconnections" is an object used to record each node used by the node group, and records the connection method of the node in the form of an array. "usage_count" stores an integer, indicating the number of times it is used. "scene" is an array used to list the scenarios in which the node group is used, including: "steelmaking scenario", "converter scenario", and "continuous casting scenario"; "tags" stores an array used to record tags related to the node, including: "planning and scheduling", "steel ladle visibility";

[0114] The function json format contains the fields "name", "type", "nodeconnections", "usage_count", "scene" and "tags". Among them, "name" stores the name of the function. The "type" attribute of the function json is "founction", indicating that the type of the object is a function. "nodeconnections" is an object used to record the nodes and node groups used by the function, and record the connection methods of the nodes and node groups in the form of an array. "usage_count" stores an integer, indicating the number of times it is used. "scene" is an array used to list the scenarios in which the function is used, including: "steelmaking scenario", "converter scenario", and "continuous casting scenario"; "tags" stores an array for recording tags related to the function, including: "planning and scheduling", "steel ladle visibility".

[0115] According to the present invention, a steelmaking and continuous casting digital twin application intelligent configuration system includes an equipment configuration module, which contains multiple submodules classified by equipment type: such as a converter module, a KR module, a LF furnace module, a RH furnace module, a continuous casting module, a crane trolley module, etc.

[0116] Each submodule provides corresponding 3D models and process action simulation for specific equipment in the steelmaking and continuous casting scenario;

[0117] Specifically, it also includes functional configuration modules to realize functions such as production planning and scheduling, material distribution optimization, production process simulation, energy consumption and emission monitoring, scene navigation, perspective switching, equipment sectioning, monitoring call and path display.

[0118] Embodiment 2, in order to cooperate with the steelmaking-continuous casting-hot rolling integrated production planning system, the steelmaking continuous casting process becomes a part of the integrated production plan. The system provided by the present invention is seamlessly integrated with other systems such as enterprise resource planning, i.e. ERP, manufacturing execution system, i.e. MES, quality management system, i.e. QMS, by calling the API interface, so that developers can easily retrieve the required data. The interface supports GET and POST request methods, multiple interfaces can be developed, and corresponding request parameters can be configured, including: system_type, specifying the external system type; query_type, specifying the requested data type and other data fields that need to be requested. The response body is in json format, and returns a data array according to the configured request parameters, which is used when developing functions.

[0119] To realize the intelligent setting of production planning and scheduling functions in the steelmaking and continuous casting digital twin application, from the developer inputting text requirements to the function implementation, such as Figure 4 As shown, there are the following steps:

[0120] Step S1: selecting a specific scene in steelmaking and continuous casting through a steelmaking and continuous casting scene template library;

[0121] The scenario template library includes continuous casting scenario, converter scenario, LF furnace scenario, RH furnace scenario, charging scenario and combination scenario. Each scenario includes corresponding equipment modules, for example, the charging scenario includes converter module, KR module and crane module. Since this embodiment is an intelligent setting of steel production planning and scheduling function, a combination scenario is selected in the scenario template library.

[0122] Step S2: After selecting a scene, modify the parameters of each device module in the scene.

[0123] In the converter module, set the converter model, rotation angle, and furnace door movement distance; in the KR module, set the KR model, track length, and slag removal range; in the LF furnace module, set the LF furnace model, electrode movement range, and argon blowing connection position. Set the remaining equipment modules in turn according to needs until the configuration of the model scene is completed.

[0124] Step S3: Inputting functional requirements through the input panel;

[0125] For example, "I have a production plan and need to see the specific scheduling of this plan in this scenario." The input text is passed to the BERT model for label extraction. After the extraction is completed, the BERT model outputs the label "Plan Scheduling" and displays it in the output panel.

[0126] Step S4: Check whether the labels output by BERT meet the initial requirements.

[0127] Specifically, check the detailed annotations of the "Planning and Scheduling" label; after checking, if the label meets the requirements, select the label to enter the settings. In this step, if there is no function label that meets the requirements, repeat step S2; if the requirements are not in the preset functions, configure them first according to the intelligent configuration method of the steelmaking and continuous casting digital twin application, and then perform subsequent operations.

[0128] Step S5: After confirming the settings, check the node groups under the "Planning and Scheduling" tab, which have different parameter settings. In the node group "Crane Position Drive", set the crane initial position, crane type, drive speed, drive data source, and drive mode; in the node group "Crane Lifting and Landing Drive", set the crane model binding, lifting animation type, landing animation type, and animation running speed. The remaining node groups have similar settings, which are set in turn according to needs.

[0129] Step S6: After the setting is completed, the production plan and scheduling functions are associated with each equipment module in the combined scene. The nodes and node groups used appear in the node intelligent optimization panel, and further logical adjustments can be made.

[0130] The details not elaborated in detail in this embodiment are logically consistent with the details that have been described in detail.

[0131] The technical contents not described in detail in this embodiment belong to the prior art known to professional and technical personnel in this field.

[0132] In other words, the present invention provides an intelligent configuration system for the development of digital twin applications for steelmaking and continuous casting, which can integrate multiple systems such as enterprise resource planning, i.e., ERP, manufacturing execution system, i.e., MES, quality management system, i.e., QMS, etc., and includes an equipment configuration module and a function configuration module, which are used to process and analyze data in the steelmaking and continuous casting production process and configure required functions. The developer inputs the text requirements, and the system extracts the relevant function tags according to the text requirements. After the developer selects the specified tag, the parameters can be adjusted and the functions can be configured, so as to achieve the purpose of rapid development of the digital twin system for steelmaking and continuous casting.

[0133] In the intelligent configuration system of the digital twin application of steelmaking and continuous casting, the equipment configuration module provides corresponding 3D models and process action simulations for each equipment in the steelmaking and continuous casting scenario. Each submodule can be freely combined according to the actual layout of the steelmaking and continuous casting plant, including: converter submodule, KR submodule, LF furnace module, RH furnace module, continuous casting submodule and trolley submodule;

[0134] The converter submodule includes multiple parameter settings such as converter model setting, angle setting, furnace door movement setting, etc.

[0135] The KR submodule includes multiple parameter settings such as KR model setting, track length setting, and slag removal action setting;

[0136] The LF furnace module includes multiple parameter settings such as LF furnace model setting, electrode movement range setting, argon blowing connection setting, etc.

[0137] The RH furnace module includes multiple parameter settings such as RH furnace model setting, tube length setting, and moving range setting;

[0138] The continuous casting submodule includes multiple parameter settings such as continuous casting machine model setting, fire cutting machine model setting, pusher machine model setting, stream number setting, billet curvature setting, etc.

[0139] The crane trolley submodule includes multiple parameter settings such as crane / trolley model setting, crane lifting height setting, and trolley lifting height setting.

[0140] In response to common requirements when developing digital twin applications for steelmaking and continuous casting, the system function configuration module is divided into specific function configuration and general function configuration.

[0141] The specific functional configuration for steelmaking and continuous casting includes the following sub-functions:

[0142] Production planning and scheduling includes multiple node groups such as crane position drive, crane lifting and lowering crane drive, trolley position drive, converter opening and closing drive, ladle / iron ladle visibility, and refining furnace motion drive;

[0143] Material distribution optimization includes multiple node groups such as molten steel position setting, position rationality detection, scrap steel position monitoring, etc.

[0144] The production process simulation includes multiple node groups such as data acquisition, billet segment drive, cutting machine motion drive, pusher motion drive, LF furnace motion drive, etc.

[0145] Energy consumption and emission monitoring includes multiple node groups such as data acquisition, trend correlation, and abnormal alarm.

[0146] The general function configuration of the steelmaking and continuous casting virtual scene includes the following sub-functions:

[0147] Scene tour includes node groups such as camera position movement, curve path definition, and movement speed setting;

[0148] View switching includes camera definition, camera position movement, rendering effect settings and other node groups;

[0149] Equipment section view includes node groups such as model section view, fixed point position modification, and display effect setting;

[0150] Monitoring calls include streaming calls, video progress control, pane dragging and other node groups;

[0151] Path display includes node groups such as curve path definition, path material effect definition, curve interpolation mode setting, etc.

[0152] By cooperating with the node groups in the above-mentioned equipment configuration module and function configuration module and modifying specific parameters, the application configuration of digital twins for different steelmaking and continuous casting scenarios can be completed.

[0153] Specifically, when the developer inputs text requirements, the present invention extracts the corresponding function labels through the BERT model. The BERT model is a pre-trained language model based on the Transformer architecture. Compared with other models, the BERT model has better semantic understanding and context processing capabilities, and can better handle downstream tasks and adapt to the complex situations in the steelmaking and continuous casting production process.

[0154] Specifically, to achieve the above effects, the BERT model needs to be trained and adjusted. The process has the following steps:

[0155] The first step is data preparation;

[0156] Prepare a labeled dataset named "Steelmaking Continuous CastingTag DataSet", abbreviated as "SCCTDS", which contains text corpus and corresponding labels of steelmaking continuous casting related requirements.

[0157] Specifically, for the application of digital twin in steelmaking and continuous casting,

[0158] Includes: sentence, tag;

[0159] The sentences include: "I need to plan and schedule all equipment", "I have a production plan that needs to be simulated", "Set 5 different perspectives facing the converter", and "I need the continuous caster to simulate the billet drawing speed and billet cutting process";

[0160] The tags include: "planning and scheduling", "perspective switching" and "production operation simulation".

[0161] The second step is data preprocessing;

[0162] Use the BertTokenizer method to perform word segmentation on the content input by the developer, and use pandas to process the data frame in the subsequent processing. Initialize a pre-trained BERT tokenizer bert-base-uncased, and use this tokenizer to convert the text processed by the BertTokenizer method into the input format of BERT. In this process, the text data is tokenized to generate the two parameters input_ids and attention_masks, which are then converted into PyTorch tensors. After the data processing is completed, merge it into a data frame;

[0163] The third step is model training;

[0164] Load the pre-trained BERT model bert-base-uncased and add a linear layer for classification. The number of output nodes of this linear layer is equal to the number of all classification labels. Use the num_tags parameter to specify the number of different labels in the classification task. Then define a PyTorch dataset class CustomDataset, which needs to implement the _len_ and _getitem_ methods so that PyTorch's DataLoader can read the data. In the _getitem_ method, a dictionary containing input_ids, attention_mask, and tags is returned for the model to use during training. Next, use the TrainingArguments class to define the training parameters. Finally, call a high-level API-Trainer for training;

[0165] Step 4: predictive reasoning;

[0166] After training is completed, you can use the fine-tuned model to input the functional requirement text for testing, and predict the corresponding function label based on the input functional requirement text. If the prediction is wrong, adjust the parameters in the above steps;

[0167] Step 5: Corresponding functional nodes;

[0168] According to the labels predicted by the BERT model, the corresponding functions are called, the function-related configuration panels are configured, and the corresponding nodes and node groups are called at the same time, and they are displayed on the smart optimization panel.

[0169] Specifically, in order to automatically record the association and usage frequency of nodes, node groups and functions, the present invention provides a user behavior detection method, which records usage information such as call time, call times, etc. each time a node, node group and function is called, so as to better analyze the usage of nodes, node groups and functions. The following is an implementation scheme for user behavior detection, which includes the following steps:

[0170] Logging step: Log records are added to each module to record the call time, number of calls and node connection status of nodes, node groups and functions.

[0171] Manual recording step: After the function is implemented, record the function label and add the corresponding BERT dataset.

[0172] Parse and modify json files: According to the log information and manually recorded function labels and data sets, the system automatically modifies the parameters in the json files of nodes, node groups and functions.

[0173] To achieve the above functions, the present invention provides a json file storage format dedicated to intelligently configured nodes, node groups and functions. This format is convenient for storing classified information and reflects the association and usage frequency between nodes, node groups and functions.

[0174] The above describes the use process and basic principles of the present invention in combination with specific embodiments. It should be pointed out that the advantages, strengths and implementation effects shown in this embodiment are only examples and not limitations. It cannot be considered that the details, advantages, strengths and implementation effects therein must be possessed by all embodiments, and it cannot be considered that the relevant details, advantages, strengths and effects that are not present in this embodiment but shown in other embodiments do not belong to the present invention.

[0175] like Figure 2 and Figure 4 As shown, this embodiment provides a process for setting the production planning and scheduling function in the steelmaking and continuous casting digital twin application, as well as a process for setting parameters by intelligently matching function labels with text requirement information input by developers after the function is configured. Figure 2 and Figure 4 The display is only an example and should not limit the function and scope of use of the present invention.

[0176] The present invention also provides a steelmaking and continuous casting digital twin application intelligent configuration system. The steelmaking and continuous casting digital twin application intelligent configuration system can be implemented by executing the process steps of the steelmaking and continuous casting digital twin application intelligent configuration method, that is, technical personnel in this field can understand the steelmaking and continuous casting digital twin application intelligent configuration method as the preferred implementation mode of the steelmaking and continuous casting digital twin application intelligent configuration system.

[0177] According to the present invention, a steelmaking and continuous casting digital twin application intelligent configuration system is provided, comprising:

[0178] Module M1: According to the production plan, the functional logic is constructed by adjusting the nodes and node groups, and then the functional parameters of the functional logic are configured, and the usage information of the nodes is recorded;

[0179] Module M2: adding the function parameters and usage information to the data set, and training the BERT model through the data set;

[0180] Module M3: Based on the information of the data set, determine whether the prediction of the BERT model is accurate. If the result is yes, execute module M4; if the result is no, re-trigger module M2;

[0181] Module M4: Determine whether the function is implemented; if the result is yes, the current log records and function parameters are recorded in the corresponding json file, the configuration is completed and the function logic is adjusted according to the configuration; if the result is no, module M1 is re-triggered.

[0182] Specifically, in the module M1, the function parameters include: function name, function usage scenario and function label;

[0183] The usage information includes the nodes used in the function implementation process, the number of node groups, function labels, node group labels and node connection methods.

[0184] Specifically, the data set is the SCCTDS data set; the SCCTDS data set is used to train the BERT model;

[0185] In the module M2, the training of the BERT model includes:

[0186] Module A1: adding data to the SCCTDS dataset; the data includes: sentence and tag;

[0187] The sentences include: "I need to plan and schedule all equipment", "I have a production plan that needs to be simulated".

[0188] The tags include: "planning and scheduling";

[0189] Module A2: word segmentation processing of the data to obtain training parameters;

[0190] Module A3: Call the Trainer class to train the BERT model through training parameters;

[0191] In the module A2, the BERT tagger bert-base-uncased is used to tag the text data in the data, thereby generating a parameter input_ids and a parameter attention_masks; the parameter input_ids and the parameter attention_masks are converted into PyTorch tensors, thereby synthesizing a PyTorch data frame;

[0192] In the module A3, based on the PyTorch data frame, the Trainer class is called to train the BERT model through training parameters.

[0193] Specifically, in the module M4, the json format of the node includes: "name", "type", "connections", "usage_count", "scene" and "tags" fields;

[0194] The "name" stores the name of the node;

[0195] The "connections" record the node names that are allowed to connect to the upstream and downstream of the node;

[0196] The "usage_count" stores an integer indicating the number of times it is used;

[0197] The "scene" is an array used to list the scenes using this node;

[0198] The "tags" storage array is used to record tags.

[0199] In the module M4, according to the input tag, the nodes and node connection methods related to the input tag function are displayed to form recommendation information one; according to the usage information, the usage frequency of the nodes is sorted to form recommendation information two; the recommendation information one and the recommendation information two are combined to obtain the total recommendation information; through the total recommendation information, the configuration is completed and the functional logic is adjusted according to the configuration.

[0200] According to a computer-readable storage medium storing a computer program provided by the present invention, it is characterized in that when the computer program is executed by a processor, the steps of the intelligent configuration method of the digital twin application of steelmaking and continuous casting are implemented.

[0201] An electronic device provided according to the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The electronic device is characterized in that when the computer program is executed by the processor, the steps of the intelligent configuration method of the digital twin application of steelmaking and continuous casting are implemented.

[0202] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.

[0203] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. An intelligent configuration method for digital twin application of steelmaking and continuous casting, characterized in that: include: Step S1: According to the production plan, construct the functional logic by adjusting the nodes and node groups, and then configure the functional parameters of the functional logic, and record the usage information of the nodes; Step S2: adding the function parameters and usage information into a data set, and training a BERT model using the data set; Step S3: judging whether the prediction of the BERT model is accurate according to the information of the data set, if the result is yes, executing step S4; if the result is no, re-executing step S2; Step S4: Determine whether the function is implemented; if the result is yes, record the current log records and function parameters in the corresponding json file, complete the configuration and adjust the function logic according to the configuration; if the result is no, re-execute step S1.

2. The intelligent configuration method for steelmaking and continuous casting digital twin application according to claim 1 is characterized in that: In the step S1, the function parameters include: function name, function usage scenario and function label; The usage information includes the nodes used in the function implementation process, the number of node groups, function labels, node group labels and node connection methods.

3. The intelligent configuration method for steelmaking and continuous casting digital twin application according to claim 2 is characterized in that: The data set is the SCCTDS data set; the SCCTDS data set is used to train the BERT model; In step S2, the training of the BERT model includes: Step A1: Add data to the SCCTDS dataset; the data includes: sentence and tag; The sentences include: "I need to plan and schedule all equipment", "I have a production plan that needs to be simulated"; The tags include: "planning and scheduling"; Step A2: word segmentation processing of the data to obtain training parameters; Step A3: calling the Trainer class to train the BERT model through training parameters; In the step A2, the text data in the data is marked using the BERT marker bert-base-uncased, thereby generating a parameter input_ids and a parameter attention_masks; the parameter input_ids and the parameter attention_masks are converted into PyTorch tensors, thereby synthesizing a PyTorch data frame; In step A3, based on the PyTorch data frame, the Trainer class is called to train the BERT model through training parameters.

4. The intelligent configuration method for steelmaking and continuous casting digital twin application according to claim 1, characterized in that: In the json file of step S4, the json format of the node includes: "name", "type", "connections", "usage_count", "scene" and "tags" fields; The "name" stores the name of the node; The "connections" record the node names that are allowed to connect to the upstream and downstream of the node; The "usage_count" stores an integer indicating the number of times it is used; The "scene" is an array used to list the scenes using this node; The "tags" storage array is used to record tags. In step S4, based on the input tag, the nodes and node connection methods related to the input tag function are displayed to form recommendation information one; based on the usage information, the usage frequency of the nodes is sorted to form recommendation information two; the recommendation information one and the recommendation information two are combined to obtain the total recommendation information; through the total recommendation information, the configuration is completed and the function logic is adjusted according to the configuration.

5. An intelligent configuration system for digital twin application of steelmaking and continuous casting, characterized in that: include: Module M1: According to the production plan, the functional logic is constructed by adjusting the nodes and node groups, and then the functional parameters of the functional logic are configured, and the usage information of the nodes is recorded; Module M2: adding the function parameters and usage information to the data set, and training the BERT model through the data set; Module M3: Based on the information of the data set, determine whether the prediction of the BERT model is accurate. If the result is yes, execute module M4; if the result is no, re-trigger module M2; Module M4: Determine whether the function is implemented; if the result is yes, the current log records and function parameters are recorded in the corresponding json file, the configuration is completed and the function logic is adjusted according to the configuration; if the result is no, module M1 is re-triggered.

6. The steelmaking and continuous casting digital twin application intelligent configuration system according to claim 1, characterized in that: In the module M1, the function parameters include: function name, function usage scenario and function label; The usage information includes the nodes used in the function implementation process, the number of node groups, function labels, node group labels and node connection methods.

7. The steelmaking and continuous casting digital twin application intelligent configuration system according to claim 1, characterized in that: The data set is the SCCTDS data set; the SCCTDS data set is used to train the BERT model; In the module M2, the training of the BERT model includes: Module A1: adding data to the SCCTDS dataset; the data includes: sentence and tag; The sentences include: "I need to plan and schedule all equipment", "I have a production plan that needs to be simulated"; The tags include: "planning and scheduling"; Module A2: word segmentation processing of the data to obtain training parameters; Module A3: Call the Trainer class to train the BERT model through training parameters; In the module A2, the BERT tagger bert-base-uncased is used to tag the text data in the data, thereby generating a parameter input_ids and a parameter attention_masks; the parameter input_ids and the parameter attention_masks are converted into PyTorch tensors, thereby synthesizing a PyTorch data frame; In the module A3, based on the PyTorch data frame, the Trainer class is called to train the BERT model through training parameters.

8. The steelmaking and continuous casting digital twin application intelligent configuration system according to claim 5, characterized in that: In the module M4, the json format of the node includes: "name", "type", "connections", "usage_count", "scene" and "tags" fields; The "name" stores the name of the node; The "connections" record the node names that are allowed to connect to the upstream and downstream of the node; The "usage_count" stores an integer indicating the number of times it is used; The "scene" is an array used to list the scenes using this node; The "tags" storage array is used to record tags. In the module M4, according to the input tag, the nodes and node connection methods related to the input tag function are displayed to form recommendation information one; according to the usage information, the usage frequency of the nodes is sorted to form recommendation information two; the recommendation information one and the recommendation information two are combined to obtain the total recommendation information; through the total recommendation information, the configuration is completed and the functional logic is adjusted according to the configuration.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the intelligent configuration method for the steelmaking and continuous casting digital twin application described in any one of claims 1 to 4 are implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the computer program is executed by a processor, the steps of the intelligent configuration method for the steelmaking and continuous casting digital twin application described in any one of claims 1 to 4 are implemented.

Images