Digital twinning-based dinitramide ammonium-based liquid propellant combustion simulation data labeling method and system

Abstract

The invention provides a dinitramide ammonium-based liquid propellant combustion simulation data labeling method and system based on digital twinning, and the method comprises the steps: constructing a labeling correlation map through combing and refining combustion physical rules and multi-batch historical labeling data of a dinitramide ammonium-based liquid propellant in different combustion environments; the method comprises the following steps: comparing new combustion simulation data generated by a digital twin system through iterative simulation with an annotation association map, extracting a covering combustion state description item and multiplexing a corresponding annotation label, synchronously separating a new combustion state description item which is not matched with the annotation association map, performing targeted annotation to obtain a newly added annotation label, and performing annotation on the newly added annotation label to obtain a new combustion state description item; and updating a combustion state description item-labeling label association corresponding record of the labeling association map, performing adaptive calibration on all combustion simulation historical labeling data generated in an early stage, updating original labeling label contents, and synchronously updating a combustion simulation data labeling reference library of the digital twin system. According to the method, the integrating degree of simulation and labeling can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data processing, in particular to a dinitramide ammonium-based liquid propellant combustion simulation data labeling method and system based on digital twinning. BACKGROUND

[0002] With the development of the integration of aerospace power technology and digital twinning technology, the combustion simulation data generated for the digital twinning system is given a standardized state identifier, which can provide a data basis for combustion performance analysis and simulation model optimization. At present, the industry generally constructs a rule base based on historical labeling data, matches new simulation data generated by digital twinning with rule items, successfully reuses the corresponding labeling label, and fails to store the data separately for processing, but this type of method does not fit the exclusive combustion operation logic of the propellant, has insufficient adaptability to special combustion scenarios, and the labeling rules are difficult to adjust in a timely manner as new combustion states appear, and logical deviations may occur in different batches of labeling data, thereby affecting the accuracy and consistency of digital twinning combustion simulation. SUMMARY

[0003] The present application provides a dinitramide ammonium-based liquid propellant combustion simulation data labeling method and system based on digital twinning.

[0004] In a first aspect, the present application provides a dinitramide ammonium-based liquid propellant combustion simulation data labeling method based on digital twinning, which comprises: refining the combustion physical laws of dinitramide ammonium-based liquid propellant under different combustion environments and multiple batches of historical labeling data, refining the fixed correspondence between various combustion state description items in the combustion process and corresponding labeling labels, constructing a labeling correlation graph containing the combustion physical constraints of such propellant, and retaining the correlation and correspondence records of the combustion state description items and labeling labels; comparing the combustion simulation new data generated by the digital twinning system through iterative simulation with the labeling correlation graph, extracting the combustion state description items covered by the labeling correlation graph and reusing the corresponding labeling labels, and simultaneously separating new combustion state description items that are not matched by the labeling correlation graph; performing targeted labeling on the separated new combustion state description items to obtain new labeling labels, integrating the new labeling labels according to the structure logic of the correlation and correspondence records, updating the combustion state description item-labeling label correlation and correspondence records of the labeling correlation graph; according to the updated labeling correlation graph, adaptively calibrating all the combustion simulation historical labeling data generated in the early stage, updating the original labeling label content, and obtaining a complete calibration labeling data set; based on the complete calibration labeling data set, synchronously updating the combustion simulation data labeling benchmark library of the digital twinning system, and completing the linkage update of the digital twinning iterative simulation process and the labeling data system.

[0005] In a second aspect, an embodiment of the present application provides a data labeling system, which comprises a processor and a memory, and the memory stores a computer program which is loaded and executed by the processor to implement the above-mentioned data labeling method for combustion simulation of ammonium dinitramide-based liquid propellant based on digital twinning.

[0006] The present application constructs a labeled correlation graph with physical constraints by combining the physical laws of such propellant combustion and historical labeling data, so that the labeling rules accurately fit the exclusive combustion logic, greatly improving the adaptability of the labeling bottom layer. By comparing the new data of digital twinning simulation with the graph, the matched item labeling is accurately reused, and the unmatched new state is separated, which not only ensures the labeling efficiency in conventional scenarios, but also avoids missing new combustion scenario requirements. After labeling the new state, the graph correlation record is updated, which can automatically adapt the labeling rules and greatly reduce the complexity of rule updating. The calibration of the full amount of historical labeling data ensures that all labeling content follows a unified labeling logic and eliminates batch bias. The synchronous update of the digital twinning system labeling benchmark library allows the latest standard to be directly used in the next round of combustion iteration simulation, improving the compatibility of simulation and labeling. BRIEF DESCRIPTION OF DRAWINGS

[0007] Figure 1 is a scene schematic diagram of an application environment provided by an embodiment of the present application; Figure 2 is a flowchart of a data labeling method for combustion simulation of ammonium dinitramide-based liquid propellant based on digital twinning provided by an embodiment of the present application; Figure 3 is a structural block diagram of a data labeling system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0008] In some embodiments, the data labeling method for combustion simulation of ammonium dinitramide-based liquid propellant based on digital twinning provided by an embodiment of the present application is applied in an application environment as shown in Figure 1 For example, as shown in Figure 1 , the application environment includes a database system 10 and a data labeling system 20. The database system 10 can be a general database, and the present application does not limit this. The data labeling system 20 can be a server, a server cluster composed of multiple servers, or a cloud computing service center.

[0009] Please refer to Figure 2 , which shows a flowchart of a data labeling method for combustion simulation of ammonium dinitramide-based liquid propellant based on digital twinning provided by an embodiment of the present application. The steps in the method can be executed by the data labeling system 20 in the above-mentioned Figure 1 . The method can include the following steps:

[0010] Step S100: combing the combustion physical laws of ammonium dinitramide-based liquid propellant under different combustion environments and multi-batch historical annotation data, refining various combustion state description items and their corresponding fixed correspondence with annotation labels in the combustion process, constructing an annotation association graph containing combustion physical constraints of the propellant, and retaining the association correspondence records of combustion state description items and annotation labels.

[0011] Ammonium dinitramide-based liquid propellant is a liquid fuel used in propulsion systems, and its combustion process is affected by various environmental factors. Combustion physical laws are the physical principles and trends followed by the propellant when it burns under different oxygen content, pressure, initial temperature, and other environmental conditions. Multi-batch historical annotation data is the data accumulated from the annotation of past propellant combustion simulation data, which contains descriptions and corresponding annotation labels under different combustion states. Combustion state description items are specific descriptions of various states in the propellant combustion process, such as combustion form, medium reaction process, and product release. Annotation labels are specific symbols or words that classify and identify these combustion states.

[0012] As an implementation, step S100 can specifically include steps S110-S160: Step S110: collect combustion process observation records of ammonium dinitramide-based liquid propellant under different oxygen content, pressure, and initial temperature environments and multi-batch historical annotation data, time-series split the observation records according to the three stages of combustion initiation, continuation, and termination, and obtain a segmented observation record set arranged in order of progress, ensuring that each segment matches the core features of the corresponding combustion stage.

[0013] Different oxygen content will affect the combustion speed and completeness of the propellant, pressure will affect the stability and reaction rate of combustion, and initial temperature will affect the ignition performance and combustion initiation time of the propellant. Combustion process observation records are data obtained by real-time monitoring of the propellant combustion process using experimental equipment, which can include temperature changes, pressure changes, flame shape, and other information. Multi-batch historical annotation data is data accumulated from the annotation of past propellant combustion simulation data, which contains descriptions and corresponding annotation labels under different combustion states.

[0014] To collect these data, a data storage system can be established to store the observations of combustion processes in different environments and the historical annotation data of multiple batches in the system. During the collection process, the data needs to be classified and organized to ensure the accuracy and integrity of the data. For example, the data can be classified according to environmental factors such as oxygen content, pressure, and initial temperature, and the observations and historical annotation data can be labeled for subsequent queries and use. The collected observations are time-sequentially split into three stages: start, continue, and end of combustion. During the splitting process, the boundaries of each stage need to be determined according to the characteristics and core features of the combustion process. For example, the start stage can be defined as the period from ignition to stable combustion of the flame, the continue stage can be defined as the period of stable combustion of the flame, and the end stage can be defined as the period from flame extinction to the end of combustion. In this way, the observation records can be split into a set of segmented observation records arranged in chronological order, ensuring that each segment matches the core features of the corresponding combustion stage.

[0015] Step S120: For each segment in the segmented observation record set, the elements are disassembled to extract three core contents: combustion form, medium reaction process, and product release situation. The extracted elements are structured and organized according to the segmentation structure to generate a set of structured description texts of combustion physical laws, and the text format is unified with the framework of the annotation labels of the multi-batch historical annotation data.

[0016] In the embodiment of the present application, the segmented observation record set is the observation record set obtained by splitting according to the combustion stages in the previous step. The combustion form is the shape, color, size, etc. of the flame when the propellant is burning, which can reflect the stability and intensity of the combustion. The medium reaction process is the chemical reaction process between the propellant and the surrounding medium (such as oxygen), including the reaction rate, consumption of reactants, and generation of products, etc. The product release situation is the information of the types and quantities of gaseous products generated during the combustion process, which can reflect the completeness of the combustion and the properties of the products.

[0017] The element disassembly of each segment in the segmented observation record set needs to use data analysis and processing technology. For example, image processing technology can be used to analyze the combustion pattern performance, extract the shape, color, size and other characteristics of the flame; chemical analysis technology can be used to monitor the medium reaction process, analyze the consumption of reactants and the generation of products; gas analysis technology can be used to detect the product release, determine the types and quantities of generated gas products. The extracted elements are structured and organized according to the segment structure, that is, the core content of the combustion pattern performance, medium reaction process and product release of each segment is sorted and classified to form a structured data structure. In order to facilitate subsequent processing and use, the format of the generated structured description text set needs to be consistent with the framework of the annotation labels of the multi-batch historical annotation data. For example, if the annotation label framework contains combustion stage, environmental factor, combustion state and other information, the structured description text set also needs to contain these information, and the format needs to be consistent.

[0018] Step S130: Each text entry in the structured description text set is matched with the annotation label of the multi-batch historical annotation data one by one, the corresponding relationship between the combination of the three core contents and the annotation label is mined, the initial corresponding list of the combustion state description item and the annotation label is generated, and the list entries are sorted according to the combustion process stage.

[0019] In the embodiment of the application, the structured description text set is the text set containing the core content of the combustion pattern performance, medium reaction process and product release generated in the previous step. The annotation label of the multi-batch historical annotation data is the label used in the past for annotating the propellant combustion simulation data. These labels are used to classify and identify different combustion states. Each text entry in the structured description text set is matched with the annotation label one by one, which needs to use a text matching algorithm. For example, a string matching algorithm can be used to compare the keywords in the text entry with the keywords in the annotation label to find the matching label. In the matching process, the semantics and context information of the text entry need to be considered to ensure the accuracy of the matching.

[0020] The corresponding relationship between the combination of the three core contents and the annotation label is mined, that is, the annotation label corresponding to the combination of different combustion pattern performance, medium reaction process and product release is analyzed. In this way, some potential rules and patterns can be found to provide a reference for subsequent annotation work. The initial corresponding list of the combustion state description item and the annotation label is generated, the matching results are sorted into a list, and the list entries are sorted according to the combustion process stage.

[0021] Step S140: Perform logical conflict checks on the initial correspondence list, compare the labels corresponding to the same core content combination under different environments, eliminate duplicate or contradictory items, adjust the order of items to be consistent with the combustion process stage, and obtain a fixed correspondence between the combustion state description items and the labels.

[0022] In this embodiment of the invention, the initial correspondence list is the list generated in the previous step that contains the correspondence between combustion state description items and labels. Logical conflict checking involves checking whether there are duplicate or contradictory entries in the list, i.e., the same core content combination corresponds to different labels under different environments, or the same core content combination corresponds to multiple different labels under the same environment. Comparing the labels corresponding to the same core content combination under different environments requires traversing and comparing the list. For example, the list can be grouped according to environmental factors, and then the entries in each group can be compared to find duplicate or contradictory entries. During the comparison process, the specific situation of the core content combination needs to be considered to ensure the accuracy of the comparison. Duplicate or contradictory entries are removed, i.e., entries with duplicates or contradictions are deleted from the list to ensure the uniqueness and accuracy of the correspondence. The order of the entries is adjusted to be consistent with the combustion process stages, i.e., the entries in the list are sorted according to the three stages of combustion start, continuation, and termination.

[0023] Step S150: Group the fixed correspondences according to the environmental type, set the combustion state description item as the starting content of the associated node, set the label as the ending content, define the connection relationship according to the combustion physics law, and generate a labeled association map containing the combustion physics constraints of dinitramide ammonium-based liquid propellant.

[0024] In one implementation, step S150 may specifically include the following steps S151 to S156: Step S151: Encode the content of each combustion state description item in the fixed correspondence between combustion state description items and labeling items. The content encoding includes three types of identifiers: combustion process stage, environment type, and core element combination, generating a unique combustion state coded text.

[0025] The fixed correspondence between the combustion state description items and the labeling is the unique correspondence obtained in the previous step. Content encoding converts the content of the combustion state description items into a predetermined coded form for easy storage and processing. The combustion process stages are the three stages of propellant combustion: initiation, continuation, and termination. Environmental type refers to various environmental factors affecting propellant combustion, such as oxygen content, pressure, and initial temperature. The core element combination is a combination of core content such as combustion morphology, medium reaction process, and product release.

[0026] Each combustion state description item needs to be coded, requiring the generation of codes based on three types of identifiers: combustion process stage, environmental type, and combination of core elements. For example, predefined characters or numbers can be used to represent each identifier, which are then combined into a coded text. To ensure the uniqueness of the codes, they need to be validated to avoid duplicate codes. Simultaneously, the code length and format need to be consistent to facilitate subsequent processing and querying.

[0027] Step S152: Encode the content of each label in the fixed correspondence. The content encoding includes three types of identifiers: label type, content category, and environment adaptation. Generate a unique label encoding text. The encoding length is completely consistent with the combustion state encoding text.

[0028] The label content in the fixed correspondence relationship corresponds to the label of the combustion state description item. Content encoding is the process of converting the label content into a predetermined coded form for easy storage and processing. Label type refers to the category of the label, such as combustion state category, product category, etc. Content classification is the specific classification of the label, such as stable combustion, unstable combustion, etc. Environmental adaptation is the environmental type to which the label applies, such as high oxygen content environment, low oxygen content environment, etc. Content encoding for each label requires generating a code based on three types of identifiers: label type, content classification, and environmental adaptation. For example, predetermined characters or numbers can be used to represent each identifier, and then they are combined into a coded text. To ensure the uniqueness of the code, it needs to be validated to avoid duplicate codes. Simultaneously, the code length must be exactly the same as the combustion state coded text for subsequent matching and association.

[0029] Step S153: Bind each combustion status code text to the corresponding label code text. The binding content includes information related to the environment type and combustion process stage. Generate a set of code binding pairs. The set of code binding pairs is sorted by both environment type and combustion process stage.

[0030] The combustion status coded text is the text obtained by encoding the combustion status description item in the previous step. The label coded text is the text obtained by encoding the label content in the previous step. Binding associates the combustion status coded text with the corresponding label coded text, forming a coded binding pair. The environment type and combustion process stage association information is the environment type and combustion process stage corresponding to the binding pair. Binding each combustion status coded text with its corresponding label coded text requires determining the binding relationship based on a fixed correspondence. For example, the label corresponding to each combustion status description item can be found by querying a fixed correspondence table, and then their coded texts can be bound. The binding content includes the environment type and combustion process stage association information, thus clarifying the applicable environment and stage for each binding pair. A set of coded binding pairs is generated, organizing all coded binding pairs into a single set. The set of coded binding pairs needs to be sorted by both environment type and combustion process stage. For example, it can be sorted first by environment type, and then sorted by combustion process stage under each environment type.

[0031] Step S154: Group the encoding binding sets according to environment type, with the combustion state encoded text as the starting node and the label encoded text as the ending node, and the environment type as the connection attribute, to construct the initial framework of the label association graph. The nodes and attributes adopt a standardized data format.

[0032] In this embodiment of the invention, the set of encoded binding pairs is the set generated in the previous step that includes binding pairs of combustion state encoded text and annotation label encoded text. The environment type refers to various environmental factors affecting propellant combustion, such as oxygen content, pressure, and initial temperature. The start node is the starting point in the annotation association graph, used to represent the combustion state encoded text. The end node is the ending point in the annotation association graph, used to represent the annotation label encoded text. The connection relationship attribute is the attribute that connects the start node and the end node; here, the environment type is used as the connection relationship attribute.

[0033] The coded binding pairs are grouped according to environmental type, classifying them based on environmental factors such as oxygen content, pressure, and initial temperature to form groups for different environments. Using the combustion state coded text as the starting node and the label coded text as the ending node, each binding pair is connected through links. The link attribute is the environmental type, clearly defining the environment to which each link applies. An initial framework for the label association graph is constructed, transforming the grouped binding pairs into a graph format that displays the relationship between the combustion state coded text and the label coded text through nodes and links.

[0034] Step S155: Optimize the initial framework of the labeled association map logically, adjust the node connection relationship to conform to the temporal logic of combustion physics, supplement the node association paths of different combustion stages under the same environment, and generate the optimized labeled association map.

[0035] In this embodiment of the invention, the initial framework of the labeled association map is the map constructed in the previous step, which includes the association between combustion state encoded text and labeled tag encoded text. Logical optimization involves adjusting the structure and connections of the map to better conform to the temporal logic of combustion physics. The temporal logic of combustion physics refers to the time sequence and logical relationships followed by the state changes and reaction processes of the propellant at different combustion stages.

[0036] Adjusting the node connections to align with the temporal logic of combustion physics requires analyzing and modifying the nodes and connections in the graph. For example, in the initial combustion stage, certain combustion states may lead to predetermined labels, while in the sustained stage, these combustion states may change, and the corresponding labels may also differ. Therefore, it is necessary to adjust the connections between nodes according to the temporal logic of combustion physics to ensure that the graph accurately reflects the actual process of propellant combustion.

[0037] Supplementing the node association paths between different combustion stages under the same environment involves identifying and adding missing association paths between different combustion stages within the same environment in the graph. For example, in a high-oxygen environment, a certain combustion state in the initial stage of combustion may transition to another combustion state in the sustained stage through a series of intermediate states. These intermediate nodes and connections need to be added to the graph to fully demonstrate the continuity of the combustion process.

[0038] Step S156: Perform integrity verification on all nodes and connections in the optimized annotation association map, remove isolated nodes or invalid connections, and generate an annotation association map containing the combustion physical constraints of dinitramide ammonium-based liquid propellant.

[0039] In this embodiment of the invention, the optimized annotation association map is the logically optimized map generated in the previous step. Integrity verification checks whether all nodes and connections in the map are complete and valid. An isolated node is a node in the map that is not connected to any other node, and an invalid connection is a connection that does not conform to the laws of combustion physics or contains logical errors.

[0040] To perform integrity verification on all nodes and connections in the optimized labeled association graph, the graph needs to be traversed and checked. For example, it can be checked whether each node has at least one incoming edge and one outgoing edge to ensure that the node is not isolated; it can also be checked whether the attributes and directions of each connection conform to the laws of combustion physics to ensure that the connection is valid.

[0041] Isolated nodes and invalid connections are removed from the graph to ensure its integrity and accuracy. Annotated association graphs containing the combustion physics constraints of dinitramide ammonium-based liquid propellants are generated, and the graph after removal is used as the final result.

[0042] Step S160: Serialize and store the starting node, ending node and connection relationship of the labeled association map, and generate a corresponding record of the combustion state description item and the label. The content of the corresponding record is consistent with the nodes and relationships of the labeled association map.

[0043] Serialization storage involves converting the start nodes, end nodes, and connections in a graph into a storable and transmittable data format, such as JSON or XML. The association record between combustion state description items and label tags is a file or database that records the correspondence between these items. Serializing the start nodes, end nodes, and connections of a label-associated graph requires using a serialization algorithm to convert the nodes and connections into a predetermined data format. For example, a JSON serialization algorithm can be used to convert the attributes and values ​​of nodes and connections into JSON strings, which can then be stored in a file or database.

[0044] Generate a record that associates the combustion state description item with the label, and store the serialized nodes and connection relationships in a file or database to form an associated record.

[0045] Step S200: Compare the new combustion simulation data generated by the digital twin system through iterative simulation with the labeled association map, extract the combustion state description items that can be covered by the labeled association map and reuse the corresponding label, and simultaneously separate the new combustion state description items that are not matched by the labeled association map.

[0046] In one implementation, step S200 may specifically include the following steps S210 to S260: Step S210: Retrieve new combustion simulation data generated by the digital twin system, and segment the new combustion simulation data into three stages: combustion start, continuous and termination. Each segment contains the combustion morphology, reaction process and product release information of the corresponding stage, generating multiple segments of combustion simulation sub-data arranged in process order. The segment boundaries are completely matched with the stage division of combustion physics.

[0047] The new combustion simulation data generated by the digital twin system is new data on the combustion of dinitramide ammonium-based liquid propellants obtained through iterative simulation. The three stages of combustion—initiation, duration, and termination—are different stages defined according to the physical laws of propellant combustion. Combustion morphology refers to the shape, color, size, and other characteristics of the flame during propellant combustion.

[0048] Retrieving new combustion simulation data generated by the digital twin system requires accessing this data from the system's storage device or database. The new combustion simulation data is segmented into three stages: combustion initiation, duration, and termination. The segmentation boundaries need to be determined based on the stage division criteria of combustion physics. For example, the initiation, duration, and termination stages can be determined based on characteristics such as temperature changes, pressure changes, and flame morphology. Each segment contains information on the combustion morphology, reaction process, and product release for the corresponding stage; that is, extracting the combustion morphology, reaction process, and product release information for each stage to form an independent sub-data set. Multiple segments of combustion simulation sub-data are generated and arranged in chronological order of combustion progress. This clearly demonstrates the entire propellant combustion process.

[0049] Step S220: Convert the format of each segment in the multi-segment combustion simulation sub-data, converting the machine-readable format into a format that is completely consistent with the natural language description of the starting node of the labeled association map, and generate multi-segment standardized combustion description text, the text content of which completely corresponds to the core information of the combustion simulation sub-data.

[0050] The format conversion of each segment in the multi-segment combustion simulation data requires the use of data conversion algorithms to transform the machine-readable format into a format completely consistent with the natural language description of the starting node of the labeled correlation graph. For example, binary data can be converted into text data, and the fields and values ​​in the data can be organized and expressed according to the natural language description of the starting node of the labeled correlation graph. Multiple standardized combustion description texts are then generated, and the converted text content is organized into multiple standardized text segments, each corresponding to one combustion simulation sub-data set.

[0051] Step S230: Compare each text entry in the multi-segment standardized combustion description text with all the starting nodes of the labeled association map one by one, compare each core element of the text entry with the starting node, generate a set of comparison results containing successful or failed matching markers, and bind the markers to the corresponding text entries one by one.

[0052] Each text entry in the standardized combustion description text is compared one by one with all the starting nodes of the labeled association map. This requires the use of text matching algorithms or similarity calculation algorithms. For example, a string matching algorithm can be used to compare the keywords in the text entries with the keywords in the starting nodes to find matching nodes. Each core element of the text entry is compared with that of the starting nodes, checking whether the core elements in the text entry, such as combustion morphology, media reaction process, and product release, are consistent with those in the starting nodes. A set of comparison results containing successful or unsuccessful matching markers is generated. Based on the comparison results, a successful or unsuccessful matching marker is added to each text entry, and the marker is bound to the corresponding text entry.

[0053] Step S240: Extract standardized combustion description text with matching success markers from the comparison result set, associate it with the combustion state description item of the corresponding starting node, retrieve the corresponding annotation label content, and triple bind the combustion simulation sub-data, combustion state description item, and annotation label content to generate a result set of reusable annotation labels.

[0054] In one implementation, step S240 may specifically include the following steps S241 to S246: Step S241: Retrieve standardized combustion description text with a successful match mark from the initial matching results of the reused label, associate it with the stage identifier and environment identifier of the corresponding combustion simulation sub-data, generate a set of successfully matched text with associated attributes, and sort the entries of the successfully matched text set according to the combustion process stage.

[0055] The standardized combustion description text with a successful match marker is the text entry marked as a successful match in the initial matching results. The stage identifier of the combustion simulation sub-data is the combustion stage to which the sub-data belongs, such as the start stage, duration stage, and termination stage. The environment identifier is the environmental type of the combustion simulation sub-data, such as a high-oxygen environment or a low-oxygen environment. Retrieving the standardized combustion description text with a successful match marker from the initial matching results using reused labels requires traversing the initial matching result set to find the text entries marked as successful matches. The stage identifier and environment identifier of the corresponding combustion simulation sub-data are then associated. By querying the metadata of the combustion simulation sub-data, the stage identifier and environment identifier of each successfully matched text entry are obtained and associated with the text entry. A set of successfully matched texts with associated attributes is generated, organizing the text entries associated with the stage identifier and environment identifier into a set. The entries in the successfully matched text set are sorted according to the combustion process stage, that is, the entries in the set are sorted according to the three stages of combustion start, duration, and termination.

[0056] Step S242: Match each text entry in the successfully matched text set with the starting node of the labeled association graph, obtain the encoded text of the combustion state description item of the corresponding starting node, bind the text entry, the encoded text of the combustion state description item, the stage identifier, and the environment identifier to generate an encoded association set, and group the encoded association set according to the environment type.

[0057] The starting node of the annotation association graph is the starting node used to represent the combustion state description item in the annotation association graph. The encoded text of the combustion state description item is the text obtained after encoding the combustion state description item. Matching each text entry in the successfully matched text set with the starting node of the annotation association graph requires using a text matching algorithm or similarity calculation algorithm to find the starting node that matches the text entry. The encoded text of the combustion state description item for the corresponding starting node is obtained by querying the annotation association graph to retrieve the encoded text of the combustion state description item for the matching starting node. The text entries, the encoded text of the combustion state description item, the stage identifier, and the environment identifier are bound together, forming a binding set. An encoded association set is generated, and all binding sets are organized into one set. The encoded association set is grouped by environment type, that is, the entries in the set are classified according to environment type, forming encoded association groups for different environments.

[0058] Step S243: For the encoded text of each combustion state description item in the encoded association set, match the encoded text of the corresponding label in the label association map, retrieve the corresponding natural language label content, bind the encoded text of the combustion state description item, the encoded text of the label, and the label content to generate a label association set.

[0059] The encoded text of the combustion state description item is the encoded text of the combustion state description item contained in each entry of the encoded association set. The encoded text of the corresponding annotation label in the annotation association map is the encoded text of the annotation label corresponding to the encoded text of the combustion state description item in the annotation association map. The natural language annotation label content is the natural language description of the annotation label.

[0060] For each combustion state description item in the coded association set, matching the coded text of the corresponding label in the label association graph requires querying the label association graph to find the coded text of the label corresponding to the coded text of the combustion state description item. Then, the corresponding natural language label content is retrieved, and the corresponding natural language label content is searched in the label association graph based on the coded text of the label.

[0061] Bind the coded text of the combustion status description item, the coded text of the label, and the label content. That is, associate and bind the coded text of the matching combustion status description item, the coded text of the label, and the label content to form a binding set.

[0062] Step S244: Bind the label content in the label association set to the corresponding combustion simulation sub-data. The binding content includes stage identifier, environment identifier, and combustion state description item. Generate a combustion simulation sub-data set with reused label. The set is sorted by both combustion process and environment type.

[0063] The annotation label content refers to the natural language annotation label content contained in each entry of the annotation label association set. Combustion simulation sub-data is the combustion simulation sub-data corresponding to the successfully matched text entry. The stage identifier is the combustion stage to which the combustion simulation sub-data belongs. The environment identifier is the environment type in which the combustion simulation sub-data is located. The combustion state description item is the combustion state description item corresponding to the combustion simulation sub-data. Binding the annotation label content in the annotation label association set to the corresponding combustion simulation sub-data requires querying the metadata of the combustion simulation sub-data and the annotation label association set to find the combustion simulation sub-data corresponding to each annotation label content, and then associating and binding them. The bound content includes the stage identifier, environment identifier, and combustion state description item, thus completely recording the relevant information of each combustion simulation sub-data. A set of combustion simulation sub-data with reusable annotation labels is generated, and all the bound combustion simulation sub-data are organized into a single set. The set is sorted in two ways: first by combustion process stage, and then by environment type within each stage.

[0064] Step S245: Perform content verification on the combustion simulation sub-data set of reused labels, compare the label content with the core information of the combustion simulation sub-data, remove mismatched items, and generate a result set of verified reused labels.

[0065] Content validation checks whether the labeled content matches the core information of the combustion simulation subdata. The core information of the combustion simulation subdata includes key aspects such as the combustion morphology, media reaction process, and product release.

[0066] Content validation of the combustion simulation subdata set with reused labels requires iterating through each entry in the set and comparing the label content with the core information of the combustion simulation subdata. For example, it can be checked whether the label accurately describes the combustion state in the combustion simulation subdata; if the label does not match the core information, the entry is considered mismatched.

[0067] Remove non-matching entries, that is, delete entries that do not match from the set to ensure that all entries in the set are accurate matches. Generate a result set of validated reusable labels, and use the set after removing non-matching entries as the final result. The data in this set can reliably reuse labels.

[0068] Step S246: Store the result set of the verified reuse label to the labeled simulation data path of the digital twin system. The stored content includes combustion simulation sub-data, combustion state description items, label content, stage identifier, and environmental identifier.

[0069] The labeled simulation data path in a digital twin system is the storage location used to store labeled simulation data within the digital twin system. Storing the verified set of reused labels in the labeled simulation data path of the digital twin system requires a data storage algorithm to store the data in the set into the specified path. The stored content includes combustion simulation sub-data, combustion state description items, label content, stage identifiers, and environmental identifiers.

[0070] Step S250: Extract standardized combustion description text with matching failure markers from the comparison result set, refine the core elements of the text, retain the combination of combustion morphology, medium reaction process, and product release, and generate new combustion state description items that are not matched by the labeled association map.

[0071] In one implementation, step S250 may specifically include the following steps S251 to S256: Step S251: Retrieve standardized combustion description texts with matching failure markers from the comparison result set, associate them with the stage identifier and environment identifier of the corresponding combustion simulation sub-data, and generate a set of failed matching texts with associated attributes. The entries in the set of failed matching texts are sorted according to the combustion process stage.

[0072] Standardized combustion description text with a failed match marker consists of text entries marked as failed matches in the comparison result set. The stage identifier of the combustion simulation sub-data is the combustion stage to which the sub-data belongs, such as the start stage, duration stage, and termination stage. The environment identifier is the environmental type of the combustion simulation sub-data, such as a high-oxygen environment or a low-oxygen environment. Retrieving the standardized combustion description text with a failed match marker from the comparison result set requires traversing the set to find the text entries marked as failed matches. The stage identifier and environment identifier of the corresponding combustion simulation sub-data are then associated. By querying the metadata of the combustion simulation sub-data, the stage identifier and environment identifier corresponding to each failed match text entry are obtained and associated with the text entry. A set of failed match texts with associated attributes is generated, organizing the text entries associated with the stage identifier and environment identifier into a set. The entries in the failed match text set are sorted according to the combustion process stage, that is, the entries in the set are sorted according to the three stages of combustion start, duration, and termination.

[0073] Step S252: Extract elements from each text entry in the failed matching text set, extracting three core contents: combustion mode, medium reaction process, and product release. Remove irrelevant or redundant information to generate a core element combination set. Each entry in the core element combination set corresponds one-to-one with the failed matching text.

[0074] Element extraction involves extracting key information from text, such as combustion morphology, reaction process, and product release. Irrelevant or redundant information refers to information in the text that is unrelated to or repeats the core elements. Element extraction for each text entry in the set of failed matches requires the use of text analysis algorithms to extract the three core categories of information: combustion morphology, reaction process, and product release. Irrelevant or redundant information is then removed through text filtering and screening to eliminate information that is irrelevant to or repeats the core elements.

[0075] Generate a set of core element combinations, organizing the refined core element combinations into a single set. Each entry in the core element combination set corresponds one-to-one with a failed match text entry, ensuring that each core element combination corresponds to a single text entry that failed to match.

[0076] Step S253: Compare each combination entry in the core element combination set with the core element combinations of all starting nodes in the labeled association map one by one. Confirm that the combination entry does not appear in any starting node, and generate an unmatched confirmation set. The unmatched confirmation set entries include combination content, stage identifier, and environment identifier.

[0077] The core element combinations of all starting nodes in the labeled association graph are the core element combinations contained in all starting nodes of the labeled association graph. Each combination entry in the core element combination set is compared one by one with the core element combinations of all starting nodes in the labeled association graph. This comparison requires using similarity calculation algorithms or text matching algorithms to check whether each combination entry matches any core element combination of any starting node in the labeled association graph. If a combination entry does not appear in any starting node, that is, if a combination entry has no matching core element combination of a starting node in the labeled association graph, then the combination entry is considered unmatched. An unmatched confirmation set is generated, which organizes the unmatched combination entries into a set. The unmatched confirmation set entries include combination content, stage identifier, and environment identifier, thus comprehensively recording the relevant information of each unmatched combination entry.

[0078] Step S254: Perform a structured description for each combination entry in the unmatched confirmation set. The description format is completely consistent with the natural language description of the starting node of the labeled association map, ensuring a complete combination containing the three core contents, and generating an initial set of new combustion state description items.

[0079] Structured description involves organizing and expressing combined entries according to a specific format, giving them a clear structure and meaning. The natural language description of the starting node in the labeled association map is the natural language description method used to represent combustion state description items. For each combined entry in the unmatched confirmation set, structured description requires organizing and expressing three core elements: combustion morphology, medium reaction process, and product release, according to the natural language description format of the starting node in the labeled association map. It is crucial to ensure a complete combination containing all three core elements; that is, the description must accurately include complete information on combustion morphology, medium reaction process, and product release. An initial set of new combustion state description items is generated by organizing the structured combined entries into a set. These new combustion state description items represent the state of the propellant under new environmental conditions or new combustion modes, requiring further annotation and processing.

[0080] Step S255: The initial set of new combustion state description items is double-sorted according to the combustion process stage and environmental type. The sorting logic is consistent with the organization logic of the annotation association map, generating an ordered set of new combustion state description items.

[0081] The combustion process stages are the initiation, continuation, and termination of propellant combustion. Environmental types refer to various environmental factors affecting propellant combustion, such as oxygen content, pressure, and initial temperature. The annotation association graph organization logic refers to the organization and sorting rules of nodes and relationships within the annotation association graph. The initial set of new combustion state description items is double-sorted according to both the combustion process stage and environmental type. First, the items in the set are sorted by combustion process stage, and then within each stage, they are sorted by environmental type. The sorting logic is consistent with the annotation association graph organization logic, ensuring that the organization of the new combustion state description item set is consistent with the annotation association graph, facilitating subsequent processing and querying.

[0082] Generate an ordered set of new combustion state descriptions, and use the sorted set as the final result. The entries in this set are arranged in order according to the combustion process stage and environmental type.

[0083] Step S256: Store the ordered set of new combustion state description items to the unlabeled dataset path of the digital twin system. The stored content includes new combustion state description items, stage identifiers, environmental identifiers, and sorting numbers.

[0084] The unlabeled dataset path in a digital twin system is the storage location used to store the data to be labeled within the digital twin system. Storing an ordered set of new combustion state descriptions to the unlabeled dataset path in the digital twin system requires a data storage algorithm to store the data from the set into the specified path. The stored content includes the new combustion state description, stage identifier, environment identifier, and sort number, thus comprehensively recording the relevant information for each new combustion state description.

[0085] Step S260: Standardize and store the result set of reused annotation labels and the new combustion state description item respectively. The storage path of the result set of reused annotation labels is consistent with the storage path of multiple batches of historical annotation data. The new combustion state description item is stored in the path of the dataset to be annotated.

[0086] The results set of reused labels and the new combustion state descriptions are stored separately in a standardized manner. This requires a data storage algorithm to store the results set of reused labels in a location consistent with the storage path of multiple batches of historical labeled data, and to store the new combustion state descriptions in the path of the dataset to be labeled. This ensures data consistency and manageability, facilitating subsequent processing and analysis of both labeled and unlabeled data.

[0087] Step S300: Targeted annotation is performed on the separated new combustion state description items to obtain new annotation labels. The new annotation labels are integrated with the structural logic of the associated records, and the combustion state description items-annotation labels associated records of the annotation association map are updated.

[0088] In one implementation, step S300 may specifically include the following steps S310 to S360: Step S310: Retrieve the ordered set of new combustion state description items, perform content parsing on each new combustion state description item, clarify the combination relationship and environmental association attributes of combustion morphology, medium reaction process, product release, and generate the parsed dataset to be labeled. The entries in the dataset to be labeled are arranged by sorting number.

[0089] Retrieving the ordered set of new combustion state descriptions requires obtaining this set from the unlabeled dataset path of the digital twin system. Content parsing of each new combustion state description requires using text analysis algorithms and domain knowledge to extract core content such as combustion morphology, media reaction process, and product release, and analyzing their combination relationships and environmental attributes. It is crucial to clarify the combination relationships and environmental attributes of combustion morphology, media reaction process, and product release; for example, determining whether a certain combustion morphology is associated with predetermined media reaction process and product release, and how this combination changes under different environmental conditions.

[0090] Step S320: Match each entry in the dataset to be labeled with similar combinations of label content in multiple batches of historical labeled data. Referencing the label framework structure of multiple batches of historical labeled data and combining the combustion physics of dinitramide ammonium-based liquid propellant, determine the labeling direction and content framework of each new combustion state description item, and generate a set of labeling frameworks.

[0091] The multiple batches of historical annotation data are accumulated from previous annotations of the propellant combustion simulation data, containing descriptions of different combustion states and corresponding annotation labels. The annotation label framework structure is the organization and format of the annotation labels used in the multiple batches of historical annotation data.

[0092] Matching each entry in the dataset to be labeled with similar combinations of labels from multiple batches of historical labeled data requires using text matching algorithms or similarity calculation algorithms to identify historical labeled data entries similar to each entry in the dataset to be labeled, and obtaining their label content. Referring to the label framework structure of multiple batches of historical labeled data, and combining it with the combustion physics of dinitramide ammonium-based liquid propellant, the label content of similar combinations is analyzed to determine the labeling direction and content framework for each new combustion state description item.

[0093] Determine the labeling direction and content framework for each new combustion state description item. For example, based on the laws of combustion physics, determine which combustion state category a new combustion state description item should belong to, and what content the corresponding label should contain.

[0094] Step S330: According to each direction and frame of the annotation framework set, generate new annotation labels for the corresponding new combustion state description items, so that the expression style, logical rules and annotation labels of multiple batches of historical annotation data are consistent, generate a new annotation label set, and the items in the new annotation label set correspond one-to-one with the dataset to be annotated.

[0095] New combustion state descriptions are entries in the dataset to be labeled. New annotation labels are new labels generated for these new combustion state descriptions. New annotation labels are generated for each direction and frame of the annotation framework set, based on the annotation direction and content framework specified for each entry in the annotation framework set, combined with the specific circumstances of the new combustion state description. The style and logical rules of the new labels must be consistent with the labels in multiple batches of historical annotation data, ensuring consistency in expression and logical relationships between the new labels and the labels in historical annotation data, facilitating subsequent unified management and use. A set of new annotation labels is generated, compiling the new annotation labels generated for each new combustion state description into a set. Each entry in the new annotation label set corresponds one-to-one with the dataset to be labeled, ensuring that each new combustion state description has a corresponding new annotation label.

[0096] Step S340: Bind the corresponding entries of the ordered set of new combustion state description items to the set of newly added labels. The binding content includes stage identifier, environment identifier, and sort number. Generate a set of binding pairs of new combustion state description items and newly added labels. The binding pairs are arranged according to the sort number.

[0097] The newly added label set is the set of newly added labels generated in the previous step that corresponds one-to-one with the dataset to be labeled. The stage identifier is the combustion stage to which the new combustion state description item belongs, such as the start stage, ongoing stage, or termination stage. The environment identifier is the environmental type in which the new combustion state description item is located, such as a high-oxygen environment or a low-oxygen environment. The sort number is the number of the new combustion state description item in the ordered set.

[0098] Binding the ordered set of new combustion state descriptions to the corresponding entries in the set of newly added labels requires querying the indices of both sets to associate and bind each new combustion state description with its corresponding newly added label. The binding content includes the stage identifier, environment identifier, and sort number, thus comprehensively recording the relevant information for each binding pair.

[0099] Step S350: Perform content validation on the set of binding pairs between the new combustion state description item and the newly added label, compare the correspondence between the core elements of the new combustion state description item and the content of the newly added label, eliminate mismatched binding pairs, and generate a validated set of binding pairs.

[0100] Content verification checks whether the correspondence between the core elements of the new combustion state description and the newly added label content is reasonable and accurate. The core elements of the new combustion state description include key information such as combustion morphology, media reaction process, and product release.

[0101] Content validation of the binding pairs between new combustion state descriptions and newly added labels requires iterating through each pair in the set. For each pair, the core elements of the new combustion state description are extracted and meticulously compared with the content of the newly added label. For example, if the new combustion state description describes the combustion mode as vigorous combustion with a certain gas as the main product, while the newly added label shows slow combustion with inconsistent product release, then this binding pair is considered mismatched. Mismatched binding pairs are removed from the set to ensure that the core elements of the new combustion state description correspond to the content of the newly added label in the remaining binding pairs. A validated binding pair set is generated, and the set after removing mismatched binding pairs is used as the final result; the binding pairs in this set are more accurate and reliable.

[0102] Step S360: Store the verified binding set to the newly added annotation temporary path in the digital twin system. The stored content includes the new combustion status description item, the newly added annotation label, the stage identifier, the environment identifier, and the sorting number.

[0103] Storing the verified binding pair set to the temporary path for newly added annotations in the digital twin system requires the use of a suitable data storage algorithm. The information for each binding pair in the set can be organized according to a specific format, such as storing it in a table, where each row represents a binding pair and the columns are the new combustion state description, new annotation label, stage identifier, environment identifier, and sorting number. Storing the organized data in the designated temporary path for newly added annotations facilitates further processing and management of the data, while also ensuring data security and traceability.

[0104] As one implementation method, step S300, which integrates the newly added annotation tags with the structural logic of the associated corresponding records, may specifically include the following steps S370~S3120: Step S370: Retrieve the associated records of the retained combustion status description items and labels, parse the core structural logic such as the grouping rules, encoding methods, and content organization order of the associated records, and generate a structural logic parsing report. The structural logic parsing report contains the complete organizational structure of the associated records.

[0105] Grouping rules are the methods by which the corresponding records are grouped according to factors such as environment type and combustion process stage to determine the correspondence between combustion state description items and labels. Encoding methods are the rules for encoding combustion state description items and labels. For example, the encoding of combustion state description items includes three types of identifiers: combustion process stage, environment type, and core element combination; the encoding of labels includes three types of identifiers: label type, content category, and environment adaptation. The content organization order is the method of arranging the various items in the corresponding records according to a certain logical order. Retrieving the retained corresponding records can be done by using the data storage interface of the digital twin system to obtain the record from its storage location. Parsing the core structural logic of the corresponding records requires detailed analysis and research. For example, by examining the data structure and metadata in the records, the grouping rules are determined; by analyzing the format and meaning of the encoded text, the encoding method is clarified; and by observing the arrangement order between items, the content organization order is identified. A structural logic parsing report is generated, compiling the parsed grouping rules, encoding methods, content organization order, and other information into a detailed report. The report content includes the complete organizational structure of the corresponding records, such as the specific content of each group, the specific encoding rules, and the logical basis for the arrangement of items.

[0106] Step S380: Encode the content of each new combustion state description item in the set after verification. The encoding rules are completely consistent with the encoding rules of the combustion state description items. It includes three types of identifiers: combustion process stage, environment type, and core element combination, and generates a unique set of new combustion state encoded text.

[0107] For each new combustion state description item in the set after verification, content encoding is performed. This requires generating coded text based on the combustion process stage, environmental type, and core element combination information within the new combustion state description item, according to predetermined encoding rules. For example, for a new combustion state description item that is in the combustion continuation stage, in a high-oxygen environment, and has a specific combustion mode and product release, this information is converted into a predetermined combination of characters or numbers according to the encoding rules to form a unique coded text. A unique set of new combustion state coded texts is generated, and the coded texts generated for each new combustion state description item are compiled into a set. To ensure the uniqueness of the encoding, the generated coded texts need to be verified to avoid duplicate encodings.

[0108] Step S390: Encode the content of each newly added annotation tag in the verified binding set. The encoding rules are completely consistent with the encoding rules of the annotation tags, including three types of identifiers: annotation type, content category, and environment adaptation, to generate a unique set of encoded text for the newly added annotation tags.

[0109] For each newly added label in the set after verification, content encoding is performed. This requires generating coded text according to predetermined encoding rules based on the label type, content category, and environmental adaptation information of the new label. For example, for a new label with a label type of combustion state, a content category of a certain combustion state, and an environmental adaptation of a high-oxygen environment, this information is converted into a predetermined combination of characters or numbers according to the encoding rules to form a unique coded text. A unique set of coded texts for each new label is generated and compiled into a set. Similarly, the generated coded texts need to be verified to ensure the uniqueness of the encoding.

[0110] Step S3100: Bind the corresponding entries of the new combustion state code text set to the newly added label code text set. The binding content includes the information related to the environment type and combustion process stage. Generate a new code binding pair set. The new code binding pair set is sorted by both environment type and combustion process stage.

[0111] Binding the new combustion state coded text set to the corresponding entries in the newly added label coded text set requires associating the corresponding coded text based on the original correspondence between the new combustion state description items and the newly added labels in the verified binding pair set. The binding content includes information related to the environment type and combustion process stage. This information is added to the binding content by querying the stage identifier and environment identifier in the original binding pair. A new coded binding pair set is generated, and the bound coded text pairs are organized into a single set. The new coded binding pair set is sorted in two ways: first, the set is grouped by environment type, and then within each environment type group, it is sorted by combustion process stage.

[0112] Step S3110: According to the grouping rules of the associated corresponding records, insert the newly added code binding pair set into the corresponding group of the associated corresponding records, adjust the order of the entries of the associated corresponding records to be consistent with the burning process stage, and generate an extended associated corresponding record containing the original and newly added correspondences.

[0113] According to the grouping rules of the associated records, the newly added set of coded binding pairs is inserted into the corresponding group of the associated records. For example, if the associated records are grouped according to environmental factors such as oxygen content, pressure, and initial temperature, then each binding pair in the newly added set of coded binding pairs is inserted into the corresponding group according to its environmental type.

[0114] Adjust the order of entries in the associated records to match the combustion process stages. After inserting the new code binding pair, reorder all entries in the associated records to ensure that the entries in each group are arranged in the order of the three stages of combustion start, duration, and termination.

[0115] An extended association record containing the existing and newly added correspondences is generated. The final result is the association record after inserting the newly added code binding pairs and adjusting their order. This extended association record integrates the existing correspondence between combustion state description items and labeling tags, as well as the newly added correspondences.

[0116] Step S3120: Perform content verification on the extended association corresponding record, compare the encoding rules, grouping rules, and content format of the new entry with the original entry, remove entries that do not conform to the rules, generate the verified extended association corresponding record, and ensure that the structure and logic of the verified extended association corresponding record are completely consistent with the association corresponding record.

[0117] Content validation of the extended association correspondence records requires traversing each entry in the extended association correspondence records and comparing new entries with existing entries in detail. For encoding rules, it checks whether the encoding of new and existing entries follows the encoding rules for combustion status description items and labels, includes the corresponding identification information, and is in the correct format. For grouping rules, it confirms whether new entries are correctly assigned to the appropriate groups according to the grouping rules of the association correspondence records. For content format, it checks whether the content structure and expression of new and existing entries are consistent. Entries that do not conform to the rules are removed. If a new entry is found to be inconsistent with the encoding rules, grouping rules, or content format, it is removed from the extended association correspondence records. A validated extended association correspondence record is generated, and the record after removing non-compliant entries is taken as the final result. It is ensured that the structure and logic of the validated extended association correspondence record are completely consistent with the original association correspondence record, thus guaranteeing the accuracy and consistency of the updated association correspondence record.

[0118] As one implementation method, step S300, updating the combustion state description item - label association corresponding record of the labeled association spectrum, may specifically include the following steps S3130~S3180: Step S3130: Retrieve the verified extended association corresponding record, convert the newly added code binding pair into a node and relationship format that can be recognized by the annotation association graph, including the starting node corresponding to the new combustion state code text, the ending node corresponding to the newly added annotation label code text, and the connection relationship attributes corresponding to the environment type, and generate a graph update data set.

[0119] The verified extended association record is retrieved from storage via the data interface of the digital twin system. The newly added coded binding pairs are converted into a node and relationship format recognizable by the annotation association graph. For each new coded binding pair, the new combustion state coded text is used as the starting node, the new annotation label coded text as the ending node, and the environment type as the connection attribute. For example, if in a new coded binding pair the new combustion state coded text is "CS001", the new annotation label coded text is "LT001", and the environment type is "high oxygen content", then "CS001" is used as the starting node, "LT001" as the ending node, and "high oxygen content" as the relationship attribute connecting the two nodes.

[0120] Generate an updated dataset for the atlas, compiling all transformed nodes and relationship information into a single set. This dataset will be used to update the labeled association atlas, ensuring it includes the correspondence between newly added combustion state descriptions and label tags.

[0121] Step S3140: Add the new start node and new end node from the updated map data set to the annotation association map, bind the environment type connection relationship attribute to the corresponding start and end nodes, and generate the initial version of the updated annotation association map.

[0122] Adding new start and end nodes from the updated graph dataset to the labeled association graph requires allocating corresponding storage space for these new nodes within the graph's data structure and integrating them with existing nodes. The environment type connection attribute is then bound to the corresponding start and end nodes. Based on the information in the updated graph dataset, a connection is established between the new start and end nodes, with the environment type used as an attribute of this connection.

[0123] Step S3150: Perform logical consistency verification on the initial version of the updated annotation association graph, check the encoding rules, grouping rules, and association logic of all nodes and connections, remove nodes and relationships with encoding errors or logical contradictions, and generate the verified annotation association graph.

[0124] Performing a logical consistency check on the initial version of the updated labeled association graph requires traversing every node and connection in the graph. For encoding rules, check whether the node encoding follows the encoding rules for combustion state descriptions and annotation labels, and whether the attributes of the connections conform to the specifications. For grouping rules, confirm whether nodes and connections are correctly assigned to the appropriate groups according to the grouping rules for the associated records. For association logic, check whether the connections between nodes are reasonable and conform to the physical laws of propellant combustion.

[0125] Remove nodes and relationships with encoding errors or logical contradictions. If a node is found to have an encoding error or a connection relationship has a logical contradiction, remove that node or relationship from the graph. Generate a validated labeled association graph, and use the graph after removing erroneous nodes and relationships as the final result.

[0126] Step S3160: Serialize and store all nodes and connections of the verified labeled association graph. The storage rules are consistent with the storage rules of the corresponding association records, and generate an updated initial version of the corresponding association records.

[0127] The verified labeled association graph requires serialization of all nodes and connections. This necessitates using a serialization algorithm to convert the nodes and connections into a storable and transmittable data format, such as JSON or XML. The storage rules must be consistent with those of the corresponding association records to ensure the stored data format and structure are identical, facilitating subsequent management and retrieval. An updated initial version of the association record is then generated by storing the serialized nodes and connections in a file or database.

[0128] Step S3170: Compare the verified labeled association map with the updated associated corresponding record of the initial version, remove entries with inconsistent content, and generate the final updated associated corresponding record of the combustion state description item and the label.

[0129] By comparing the verified annotation-based correlation map with the updated correlation-based records (initial version), the nodes and connections in the map need to be compared one by one with the entries in the records. Each entry in the record must be checked to ensure it has a corresponding node and connection in the map and that the information is consistent. If an entry is found to be missing a corresponding node in the map, or if the entry information is inconsistent with the map information, that entry is removed from the record. The final updated correlation-based records of combustion state descriptions and annotations are then generated, and the records after removing inconsistent entries are taken as the final result.

[0130] Step S3180: Overwrite the original associated records with the final updated combustion state description item and the associated records with the labels, overwrite the original labeled associated records with the verified labeled associated maps, and synchronously store them in the specified path of the digital twin system.

[0131] The updated combustion state description and label association records are overwritten with the original records. Data storage operations are then performed to store the updated records in the same location as the original records, replacing the old ones. Similarly, the verified label association map is overwritten with the original map, and the updated map is stored in the same location as the original map. All data is synchronously stored in the specified path within the digital twin system, ensuring consistency in the storage locations of the association records and the label association map, facilitating unified management and use of this data. By overwriting the original records and maps, the label association map and its corresponding records are updated, reflecting the latest correspondence between the combustion state description and the labels.

[0132] Step S400: Based on the updated annotation association map, perform adaptive calibration on all previously generated historical combustion simulation annotation data, update the original annotation label content, and obtain a complete calibration annotation dataset.

[0133] As one implementation method, step S400 involves adaptively calibrating all previously generated historical combustion simulation annotation data based on the updated annotation association map. This may specifically include the following steps S410~S460: Step S410: Retrieve the updated annotation association map, extract the new start node, new end node, and new connection relationship attributes, generate a new corresponding relationship set, and group the new corresponding relationship set according to environment type and combustion process stage, with the grouping rules being consistent with the annotation association map.

[0134] Retrieve the updated annotation-related graph from storage via the data interface of the digital twin system. Extract the newly added start nodes, newly added end nodes, and newly added connection attribute. This requires comparing the updated annotation-related graph with the original one to identify the newly added nodes and connection attribute. Generate a set of newly added correspondences, organizing the newly added start nodes, end nodes, and connection attribute into a set, recording the correspondence between the newly added combustion state description items and annotation labels. The set of newly added correspondences is grouped by environment type and combustion process stage, with grouping rules consistent with the annotation-related graph.

[0135] Step S420: Retrieve all previously generated historical annotation data of combustion simulation, perform dual classification according to combustion process stage and environmental type, and generate a multi-classified historical annotation dataset. The entries in the multi-classified historical annotation dataset are sorted by stage and environment, and the sorting logic is consistent with the annotation association map.

[0136] The combustion process consists of three stages: the start, continuation, and termination of propellant combustion. Environmental type refers to various environmental factors affecting propellant combustion, such as oxygen content, pressure, and initial temperature. All previously generated historical combustion simulation annotation data is retrieved from storage via the digital twin system's data interface. A dual classification system is implemented based on both combustion process stage and environmental type. Each entry in the historical annotation data is categorized according to its combustion process stage and environmental type; for example, entries in the initial combustion stage under high oxygen content conditions are grouped together. A multi-classified historical annotation dataset is generated, and the categorized historical annotation data is organized into a single dataset. Entries in the multi-classified historical annotation dataset are sorted by stage and environment, following the same sorting logic as the annotation association map: first sorted by combustion process stage, and then sorted by environmental type within each stage.

[0137] Step S430: Compare each entry in the multi-class historical annotation dataset with the content of the newly added starting node in the newly added correspondence set one by one, identify the combustion state description item that perfectly matches the newly added starting node, and generate a set of historical annotation data entries with calibration marks.

[0138] Each entry in the multi-class historical annotation dataset is compared one by one with the content of the newly added starting node in the newly added correspondence set. This requires using text matching algorithms or similarity calculation algorithms to compare the combustion state descriptions in the historical annotation data entries with the content of the newly added starting nodes. Combustion state descriptions that perfectly match the newly added starting nodes are identified. If a combustion state description in a historical annotation data entry is completely identical to the content of a newly added starting node, that entry is considered to need calibration. A set of historical annotation data entries with calibration tags is generated, and calibration tags are added to the historical annotation data entries that need calibration. These tagged entries are then compiled into a single set.

[0139] Step S440: Separate the set of historical labeled data entries with calibration marks from the multi-class historical labeled dataset, store them separately from the entries without calibration marks, and generate a historical labeled dataset to be calibrated and an uncalibrated historical labeled dataset. The storage paths of the two datasets are independent of each other.

[0140] The set of historical labeled data entries with calibration marks is separated from the multi-class historical labeled dataset. Through data filtering operations, the entries with calibration marks are extracted from the multi-class historical labeled dataset. They are stored separately from the entries without calibration marks. The entries with calibration marks are stored in a separate storage path to form the historical labeled dataset to be calibrated; the entries without calibration marks are stored in another separate storage path to form the uncalibrated historical labeled dataset.

[0141] Generate a historical annotation dataset to be calibrated and an uncalibrated historical annotation dataset. The two datasets are stored in independent paths, which facilitates different processing of the two datasets later. The historical annotation dataset to be calibrated will have its annotation labels updated, while the uncalibrated historical annotation dataset can remain unchanged or undergo other processing.

[0142] Step S450: For each entry in the historical annotation dataset to be calibrated, match the newly added starting node with the newly added ending node content of the newly added corresponding relationship set, i.e., the newly added annotation label content, and bind the entry to be calibrated with the newly added annotation label to generate a calibration data set to be processed.

[0143] The newly added start node is the start node representing the combustion state description item in the newly added correspondence set, and the newly added end node content is the specific content of the end node representing the label in the newly added correspondence set, that is, the newly added label content.

[0144] For each entry in the historical annotation dataset to be calibrated, the content of the newly added terminal node in the newly added correspondence set is matched with the corresponding newly added start node. For each entry in the historical annotation dataset to be calibrated, the corresponding newly added start node is found based on its combustion state description item. Then, the content of the corresponding newly added terminal node, i.e. the newly added annotation label content, is found in the newly added correspondence set through the newly added start node.

[0145] Bind the items to be calibrated to the newly added labels, and associate and bind information such as combustion status descriptions, stage identifiers, and environmental identifiers in the historical labeled data items to be calibrated with the newly added labels. Generate a set of calibration data to be processed, and organize the bound items into a set.

[0146] Step S460: Store the set of calibration data to be processed to the calibration temporary path of the digital twin system. The stored content includes the items to be calibrated, the content of the newly added start node, the newly added label, and the calibration mark. The storage format is consistent with the historical label data of multiple batches.

[0147] Storing the calibration dataset to be processed in the digital twin system's temporary calibration path requires using a data storage algorithm to store the data in the dataset to the specified path. The stored content includes the calibration entries, newly added start node information, newly added annotation labels, and calibration markers, comprehensively recording the relevant information for each calibration entry and the annotation labels that need updating. The storage format is consistent with multiple batches of historical annotation data, ensuring data consistency and manageability, facilitating subsequent processing and analysis of the calibration data.

[0148] As one implementation method, step S400, updating the original annotation label content to obtain a complete calibration annotation dataset, may specifically include the following steps S470~S4180: Step S470: Retrieve the calibration data set to be processed, separate the historical annotation entries to be calibrated, the content of newly added start nodes, the content of newly added annotation labels, and the calibration marks, and generate a detailed set of entries to be calibrated. The entries in the detailed set of entries to be calibrated are arranged according to their original sorting numbers.

[0149] The "Historical Label Entries to be Calibrated" are historical labeled data entries in the dataset that require label updates. The "New Start Node Content" is the specific content of the new start node corresponding to the historical label entries to be calibrated. The "New Label Tag Content" is the new label that needs to replace the original label of the historical label entries to be calibrated. The "Calibration Mark" is a mark added to the historical label entries to be calibrated to identify that calibration is required.

[0150] The calibration data set to be processed is retrieved from the temporary calibration path via the data interface of the digital twin system. The historical annotation entries to be calibrated, the content of newly added start nodes, the content of newly added annotation labels, and the calibration marks need to be separated. This requires parsing each entry in the calibration data set to extract the historical annotation entries to be calibrated, the content of newly added start nodes, the content of newly added annotation labels, and the calibration marks.

[0151] Generate a detailed set of items to be calibrated, and organize the separated historical annotation items into a single set. The items in the detailed set of items to be calibrated are arranged according to their original sorting numbers, maintaining the original sorting of the historical annotation data items.

[0152] Step S470: Retrieve the calibration data set to be processed, separate the historical annotation entries to be calibrated, the content of newly added start nodes, the content of newly added annotation labels, and the calibration marks, and generate a detailed set of entries to be calibrated. The entries in the detailed set of entries to be calibrated are arranged according to their original sorting numbers.

[0153] The historical annotation entries to be calibrated are the original historical annotation data entries contained in the calibration dataset. These entries require updated annotation labels due to changes in the correspondence between combustion states and the updated annotation association map. The newly added start node content is the specific description of the start node representing the combustion state description item in the newly added correspondence in the updated annotation association map, used to match the combustion state description items in the historical annotation entries to be calibrated. The newly added annotation label content is determined based on the new correspondence and replaces the original annotation labels in the historical annotation entries to be calibrated. The calibration mark is a special identifier previously added to these entries requiring calibration for easy identification and differentiation.

[0154] The calibration dataset is retrieved from the temporary calibration path using the data reading function of the digital twin system. Historical annotation entries, newly added start node content, newly added annotation label content, and calibration marks are separated. Data parsing algorithms can be used to analyze each entry in the calibration dataset in detail. For example, if the calibration dataset is stored in a table format, with each row representing an entry and different columns storing information such as historical annotation entries, newly added start node content, newly added annotation label content, and calibration marks, then these different types of content can be separated by extracting data by column. A detailed set of calibration entries is generated, summarizing the separated historical annotation entries into a new set. The entries in the detailed set are arranged according to their original sorting numbers to maintain the original order of the historical annotation data and ensure the continuity and traceability of data in subsequent processing. For example, in the original historical annotation dataset, each entry has a unique sorting number; when generating the detailed set of calibration entries, the entries are sorted according to this number.

[0155] Step S480: For each item to be calibrated in the detailed set of items to be calibrated, replace the original label content with the corresponding newly added label content. The replacement process retains the combustion state description, stage identifier, and environmental identifier of the item to be calibrated, and generates an updated set of historical label data items.

[0156] When processing each item in the detailed set of items to be calibrated, the original label content is first located. Then, it is precisely replaced with the corresponding new label content. During the replacement process, special attention should be paid to preserving the combustion state description, stage identifier, and environmental identifier of the item to be calibrated. This is because this information is a key element describing the combustion situation represented by the item and is of great significance for subsequent data analysis and simulation. For example, the combustion state description details the specific state of propellant combustion, the stage identifier clarifies whether the combustion is in the initial, ongoing, or terminated stage, and the environmental identifier indicates the specific environmental conditions under which combustion occurs, such as oxygen content and pressure. An updated set of historical labeled data items is then generated, summarizing all items to be calibrated after the label replacement operation into a new set. The items in this set have been updated with labels, more accurately reflecting the correspondence between the propellant combustion state and the labels.

[0157] Step S490: Perform content verification on the updated set of historical annotation data entries, compare the correspondence between the updated annotation label content and the combustion state description item and the newly added start node content, remove mismatched updated entries, and generate a verified set of updated historical annotation entries.

[0158] Comparing the updated label content with the combustion state description and the newly added start node content requires a detailed analysis of each entry in the updated historical label data set. For example, it's necessary to check whether the combustion state described by the updated label matches the core elements in the combustion state description, such as the actual combustion mode, medium reaction process, and product release. Simultaneously, it's also crucial to verify consistency with the combustion state represented by the newly added start node content. If a mismatch is found—for example, if the label describes slow combustion while the combustion state description shows vigorous combustion—this entry is considered a mismatched update entry. These mismatched update entries are removed from the updated historical label data set. A validated updated historical label entry set is then generated. This filtering process ensures that the entries in the resulting set have a more accurate and reliable correspondence between the label, combustion state description, and newly added start node content.

[0159] Step S4100: Merge the verified set of updated historical annotation entries with the uncalibrated historical annotation dataset. The merging process is sorted by both combustion process stage and environment type. The sorting logic is consistent with that of the multi-class historical annotation dataset, generating a complete set of merged historical annotation data.

[0160] Merging these two datasets requires a data merging algorithm. During the merging process, the data is sorted by both combustion process stage and environmental type. First, based on the combustion process stage, the data is divided into three main categories: start, ongoing, and terminated. Then, within each stage category, it is further subdivided according to environmental type, such as different oxygen contents, pressures, and initial temperatures. The sorting logic is consistent with the multi-class historical annotation dataset to ensure a unified structure and order in the merged dataset. A complete merged historical annotation dataset is generated by merging the validated updated historical annotation entries and the uncalibrated historical annotation dataset according to the above sorting method into a single complete dataset. This dataset contains all calibrated and uncalibrated historical annotation data.

[0161] Step S4110: Perform structural verification on the merged historical annotation data set to ensure that the data structure and format of all entries are completely consistent with the historical annotation data from multiple batches, remove entries with duplicate content or logical contradictions, and generate the structurally verified historical annotation data set.

[0162] To ensure that the data structure and format of all entries are completely consistent with the historical annotation data from multiple batches, each entry in the merged historical annotation data set needs to be checked. For example, check whether each entry contains the same fields, such as combustion status description, annotation labels, stage identifiers, environmental identifiers, etc., and whether the data type and format of these fields are consistent with the historical annotation data from multiple batches. If an entry is found to be missing necessary fields or the field format does not meet the requirements, then that entry needs to be adjusted or removed.

[0163] Remove entries with duplicate content or logical contradictions. By comparing all entries in the merged historical annotation dataset, identify duplicate entries with identical content and entries with logical contradictions, such as contradictory labels appearing in the same environment and stage, and remove these entries from the dataset.

[0164] The generated historical labeled data set is structure-verified. After structure verification and elimination operations, the resulting dataset is more standardized in data structure and format, and the content is more accurate.

[0165] Step S4120: Store the complete set of historical annotation data after structural verification to the temporary historical annotation path of the digital twin system. The stored content includes all historical annotation entries, updated annotation labels, stage identifiers, and environment identifiers. The storage format is consistent with that of multiple batches of historical annotation data.

[0166] The stored content includes all historical annotation entries, updated annotation labels, stage identifiers, and environment identifiers. This information comprehensively records the relevant details of each historical annotation entry. The storage format is consistent with multiple batches of historical annotation data to ensure data consistency and compatibility, facilitating unified querying, analysis, and use of the data subsequently. For example, if multiple batches of historical annotation data use a predetermined file format (such as CSV, JSON, etc.), then the same format will be used when storing the complete set of historical annotation data after storage structure verification. This avoids data processing difficulties caused by inconsistent formats.

[0167] Step S4130: Retrieve the complete set of historical annotation data after structural verification, perform logical consistency verification on the content of the complete set, compare the annotation label content of the same environment and the same stage to ensure that the annotation label content is logically consistent and has no contradictory expressions, and generate the complete set of historical annotation data after logical verification.

[0168] By comparing the label content of entries in the same environment and stage, the entire set of historical labeled data after structural verification is compared one by one for entries in the same environment and combustion stage. For example, in the combustion duration stage under high oxygen content environment, the label content of all entries is checked to see if they are consistent and if there are any contradictory statements. If a label for an entry is found to conflict with other entries in this environment and stage, such as one labeling it as stable combustion and another as unstable combustion, then these labels need to be further verified and corrected. Ensuring logical consistency and no contradictory statements in the label content, through such comparison and correction, guarantees the logical consistency of the labels in the dataset under the same environment and stage. A complete set of historical labeled data after logical verification is generated. The dataset obtained after logical consistency verification and correction is logically more rigorous.

[0169] Step S4140: Standardize the format of the complete set of historical annotation data after logical verification to ensure that the field name, order, and format of each entry are completely consistent with the format of the combustion simulation data annotation benchmark library of the digital twin system, and generate a complete set of standardized historical annotation data.

[0170] To ensure that the field names, order, and format of each entry are completely consistent with the format of the combustion simulation data annotation benchmark library of the digital twin system, the format of each entry in the complete historical annotation data set after logical verification needs to be adjusted. For example, if the field names in the benchmark library use a predetermined naming convention, while the field names in the dataset are different, then the field names need to be modified; if the field order in the benchmark library has specific requirements, while the order in the dataset is inconsistent, then the field order needs to be rearranged; if there are differences in the field format (such as date format, numerical precision, etc.), corresponding conversions also need to be performed. A standardized complete historical annotation data set is generated. After this standardization format processing, the resulting dataset is completely consistent in format with the combustion simulation data annotation benchmark library of the digital twin system.

[0171] Step S4150: Perform double sorting on the complete set of standardized historical annotation data according to the combustion process stage and environmental type. The sorting logic is consistent with the organization logic of the annotation association map, generating an ordered historical annotation dataset. The entries in the ordered historical annotation dataset are numbered sequentially.

[0172] The standardized historical annotation data set is sorted in two ways: by combustion process stage and by environmental type. First, the data is divided into three categories based on the combustion process stage: start, ongoing, and terminated. Then, within each stage category, it is further subdivided according to environmental type. The sorting logic is consistent with the organization logic of the annotation association map. This is because the organization method of the annotation association map has fully considered the influence of combustion process and environmental factors on combustion state. Sorting the standardized historical annotation data set according to the same logic makes the data organization more reasonable and facilitates correlation and comparison with the annotation association map.

[0173] An ordered historical annotation dataset is generated, resulting in a dataset with a clear structure and order after sorting. The entries in the ordered historical annotation dataset are numbered sequentially, with each entry assigned a unique sequential number, thus improving data processing efficiency.

[0174] Step S4160: Perform content duplication checks on the ordered historical annotation dataset, compare the content of all entries, remove completely duplicate entries, and generate a non-duplicate ordered historical annotation dataset to ensure that the content of the non-duplicate ordered historical annotation dataset is concise and accurate.

[0175] Compare the content of all entries by performing pairwise comparisons on each entry in the ordered historical annotation dataset. Data comparison algorithms can be used to compare key information such as combustion state descriptions, annotation labels, stage identifiers, and environmental identifiers. If two entries are found to be identical in all key information, they are considered duplicates. Completely duplicate entries are then removed from the ordered historical annotation dataset. This process generates a unique, ordered historical annotation dataset. The resulting dataset is more concise and accurate, avoids data redundancy, and improves data quality and usability.

[0176] Step S4170: Perform final content verification on the non-repeating ordered historical annotation dataset to confirm that all entries are accurate, formatted uniformly, and logically consistent. After verification, store the data in the complete calibration dataset path of the digital twin system to generate a complete calibration annotation dataset.

[0177] All entries are verified to be accurate, formatted consistently, and logically coherent. A comprehensive check is performed on each entry in the non-duplicate, ordered historical annotation dataset. The check includes verifying that the combustion state description accurately reflects the propellant's combustion state, that the annotation labels match the combustion state description, that field formats meet requirements, and that logical relationships are reasonable. For example, the check ensures that the annotation labels are consistent with core elements in the combustion state description, such as combustion morphology, media reaction process, and product release, and that field numerical precision and date formats meet standards. After successful verification, the dataset is stored in the complete calibration dataset path of the digital twin system. This process generates a complete calibration annotation dataset. The resulting dataset achieves a high quality standard in terms of content, format, and logic.

[0178] Step S4180: Synchronize the complete calibration annotation dataset to the designated shared path of the digital twin system to provide data support for the subsequent update of the combustion simulation data annotation benchmark library. After synchronization is completed, generate data synchronization confirmation information to ensure that subsequent operations can directly retrieve the complete calibration annotation dataset.

[0179] The complete calibration label dataset is synchronized to a designated shared path in the digital twin system. A data synchronization algorithm copies the complete calibration label dataset from the complete calibration dataset path to the designated shared path. During synchronization, data integrity and accuracy must be ensured to prevent data loss or corruption. This provides data support for subsequent updates to the combustion simulation data labeling benchmark library. The designated shared path is a public data storage location; subsequent updates to the combustion simulation data labeling benchmark library can directly retrieve the complete calibration label dataset from this path, ensuring the benchmark library promptly reflects the latest combustion status and the correspondence between labeling tags. Upon completion of synchronization, a data synchronization confirmation message is generated. This message can be a notification or a predefined identifier file. This ensures that subsequent operations can directly retrieve the complete calibration label dataset. By generating the data synchronization confirmation message, subsequent updates to the combustion simulation data labeling benchmark library can confirm successful data synchronization, allowing direct retrieval of the complete calibration label dataset from the designated shared path for benchmark library updates, improving data processing efficiency and reliability.

[0180] Step S500: Based on the complete calibration and annotation dataset, synchronously update the combustion simulation data annotation benchmark library of the digital twin system to complete the linkage update of the digital twin iterative simulation process and the annotation data system.

[0181] In one implementation, step S500 may specifically include the following steps S510-S560: Step S510: Retrieve the complete calibration annotation dataset, perform dual classification according to combustion process stage and environmental type, and generate a multi-class calibration data set. The classification rules of the multi-class calibration data set are completely consistent with the combustion simulation data annotation benchmark library of the digital twin system.

[0182] The complete calibration annotation dataset is stored in a designated path within the digital twin system, and can be retrieved through the system's data reading interface. The dual classification by combustion process stage and environmental type is designed to better match and integrate the calibration data with the combustion simulation data annotation benchmark library.

[0183] The combustion process is divided into three stages: initiation, continuation, and termination. Environmental types include varying oxygen content, pressure, initial temperature, and other factors. During classification, each entry in the complete calibration dataset is categorized according to its combustion process stage and environmental type. For example, entries in the initial combustion stage under high oxygen content conditions are grouped together.

[0184] A multi-class calibration dataset is generated. After classification, all categorized items are organized into a single set. The classification rules of the multi-class calibration dataset are completely consistent with the combustion simulation data annotation benchmark library of the digital twin system. This ensures that when the benchmark library is updated subsequently, the calibration data can accurately correspond to the corresponding position in the benchmark library, facilitating data replacement and updates.

[0185] Step S520: Convert the multi-class calibration data set into a storage format that can be directly imported from the combustion simulation data annotation benchmark library, ensuring that the field names, order, and format of each category of data are completely consistent with the corresponding category in the combustion simulation data annotation benchmark library, and generate a standardized calibration data import package.

[0186] Although the multi-class calibration dataset has been classified according to the same classification rules as the combustion simulation data annotation benchmark library, it may be incompatible with the benchmark library in terms of storage format. Format conversion is intended to solve this problem.

[0187] Converting the format of a multi-class calibration dataset requires the use of a data format conversion algorithm. First, the storage format supported by the combustion simulation data annotation benchmark library must be determined; this could be a predefined database format (such as SQLite, MySQL, etc.) or file format (such as CSV, JSON, etc.). Then, based on this target format, the data for each category in the multi-class calibration dataset is adjusted.

[0188] Ensure that the field names, order, and format of each category data are completely consistent with the corresponding category in the combustion simulation data annotation benchmark library. Check whether the field names in the category data are the same as those in the benchmark library; if they differ, modify them. Adjust the field order to match the benchmark library. Convert the field formats (such as data type, numerical precision, date format, etc.) to meet the requirements of the benchmark library. Generate a standardized calibration data import package. After format conversion, package all category data into a standardized import package. This import package can be directly imported into the combustion simulation data annotation benchmark library.

[0189] Step S530: Upload the standardized calibration data import package to the specified import path of the combustion simulation data annotation benchmark library. During the upload process, verify the integrity and format correctness of the standardized calibration data import package. After the upload is completed, generate upload confirmation information.

[0190] The standardized calibration data import package is a data package that has undergone format conversion and can be used to update the benchmark library. The designated import path for the combustion simulation data annotation benchmark library is a storage location specifically set up in the digital twin system for importing calibration data. Uploading the standardized calibration data import package to the designated import path can be done using data upload protocols (such as FTP, HTTP, etc.). During the upload process, the integrity and format correctness of the standardized calibration data import package are verified. Integrity verification is achieved by calculating the hash value of the import package (such as MD5, SHA-1, etc.) and comparing it with the hash value of the original data. If the hash values ​​are the same, it means that no data was lost or corrupted during the transmission of the import package. Format correctness verification checks whether the file format of the import package conforms to the requirements of the combustion simulation data annotation benchmark library, such as checking whether the file extension and file structure are correct. An upload confirmation message is generated after the upload is completed. The system automatically generates an upload confirmation message when the import package is successfully uploaded to the designated import path.

[0191] Step S540: Start the combustion simulation data annotation benchmark library import program of the digital twin system, import the multi-class calibration data in the standardized calibration data import package into the combustion simulation data annotation benchmark library, overwrite the original annotation label data of the corresponding environment and stage, and record the number of data entries covered during the import process.

[0192] Initiate the import program by triggering the combustion simulation data annotation benchmark library import process through the digital twin system's user interface or command-line tool. The program imports multi-category calibration data from the standardized calibration data import package into the combustion simulation data annotation benchmark library. It reads the data from the import package and writes it into the benchmark library. During the import process, the calibration data overwrites the original annotation labels for the corresponding environment and stage based on the combustion process stage and environmental type. For example, calibration data from the continuous combustion stage under low oxygen conditions replaces the original annotation data for that environment and stage in the benchmark library. The import process records the number of data entries covered. This information can be used for subsequent data analysis and monitoring to understand the impact of the calibration data on the benchmark library.

[0193] Step S550: Start the combustion simulation data annotation benchmark library verification program to verify the overall logical consistency between the imported calibration data and the combustion simulation data annotation benchmark library, confirm that there are no content conflicts, format errors, or data omissions, and generate verification confirmation information after the verification is passed.

[0194] The Combustion Simulation Data Labeling Benchmark Library Verification Program is a software program used to check the accuracy and consistency of data in the benchmark library. Although the imported calibration data has undergone previous format conversion and upload verification, conflicts or inconsistencies may still arise with the original data after importing into the benchmark library. Start the Combustion Simulation Data Labeling Benchmark Library Verification Program through the digital twin system's interface or command-line tool. The program verifies the overall logical consistency between the imported calibration data and the combustion simulation data labeling benchmark library. It performs a comprehensive check on all data in the benchmark library. The checks include whether the logical relationships between the data are reasonable (e.g., whether the labeling tags match the combustion state descriptions); whether the format is correct (e.g., whether the field type, length, and precision meet requirements); and whether there are any missing data (e.g., whether labeling data is missing for a particular environment or stage). It confirms there are no content conflicts, format errors, or data omissions. If any problems are found during the verification process, the program will record these issues and take appropriate action, such as prompting the user to make corrections. If all checks pass, it means that the imported calibration data is logically consistent with the benchmark library and there are no content conflicts, format errors, or data omissions.

[0195] Step S560: After the verification is passed, start the next round of combustion iteration simulation of dinitramide ammonium-based liquid propellant in the digital twin system, use the updated combustion simulation data annotation benchmark library as the annotation basis for the simulation, generate linkage update completion confirmation information after the simulation starts, and complete the linkage update of the digital twin iteration simulation process and the annotation data system.

[0196] Once the combustion simulation data annotation benchmark library passes verification, it indicates that the benchmark library has been successfully updated and the data quality is reliable. At this point, the next round of iterative simulation of dinitramide ammonium-based liquid propellant combustion in the digital twin system can be initiated.

[0197] The next round of iterative combustion simulation of dinitramide ammonium-based liquid propellant is initiated using the digital twin system. Initial conditions and parameters are set via the simulation control module, and the simulation program is then launched. The updated combustion simulation data annotation benchmark library is used as the basis for simulation annotation. During the simulation, the digital twin system annotates the simulated combustion data according to the updated benchmark library, ensuring the accuracy and timeliness of the annotation information. A confirmation message indicating the completion of the linkage update is generated upon successful simulation launch. This message signifies that the linkage update between the digital twin iterative simulation process and the annotation data system has been completed. Specifically, by updating the benchmark library, the annotation data system can reflect the latest combustion state and the correspondence between annotation labels, and the updated annotation data can be used for annotation in the new iterative simulation.

[0198] Please refer to Figure 3 This diagram illustrates the structure of a data annotation system 20 provided in one embodiment of the present invention. This data annotation system can be used to implement the functions of the aforementioned data annotation method for combustion simulation of dinitramide ammonium-based liquid propellants based on digital twins. Specifically:

[0199] The data annotation system 20 includes a central processing unit (CPU) 21, a system memory 24 including random access memory (RAM) 22 and read-only memory (ROM) 23, and a system bus 25 connecting the system memory 24 and the CPU 21. The data annotation system 20 also includes a basic input / output system (I / O system) 26 to facilitate information transfer between various devices within the computer, and a mass storage device 27 for storing the operating system 271.

[0200] The input / output system 26 may include a display for showing information and input devices such as a mouse and keyboard for user input. Both the display and the input devices are connected to the central processing unit 21 via an input / output controller connected to the system bus 25.

[0201] Mass storage device 27 is connected to central processing unit 21 via a mass storage controller (not shown) connected to system bus 25. Mass storage device 27 and its associated computer-readable media provide non-volatile storage for data labeling system 20. That is, mass storage device 27 may include computer-readable media (not shown) such as hard disk or CD-ROM (Compact Disc Read-Only Memory) drive.

[0202] According to various embodiments of the present invention, the data annotation system 20 can also be connected to a remote computer on a network such as the Internet. That is, the data annotation system 20 can be connected to a network 29 via a network interface unit 28 connected to the system bus 25, or the network interface unit 28 can be used to connect to other types of networks or remote computer systems (not shown). The memory also includes a computer program stored in the memory and configured to be executed by one or more processors to implement the above-described method.

Claims

1. A method for labeling combustion simulation data of dinitramide ammonium-based liquid propellants based on digital twins, characterized in that, The method includes: We analyzed and refined the combustion physics of dinitramide ammonium-based liquid propellants under different combustion environments and historical annotation data from multiple batches. We extracted the fixed correspondence between various combustion state description items and corresponding annotation labels during the combustion process, constructed an annotation association map containing combustion physics constraints of this type of propellant, and retained the association correspondence records between combustion state description items and annotation labels. The new combustion simulation data generated by the digital twin system through iterative simulation is compared with the labeled association map. Combustion state description items that can be covered by the labeled association map are extracted and the corresponding labels are reused. New combustion state description items that are not matched by the labeled association map are separated simultaneously. The newly separated combustion state description items are labeled in a targeted manner to obtain new label tags. The new label tags are integrated with the structural logic of the associated corresponding records to update the combustion state description item-label associated corresponding records of the label association map. Based on the updated annotation association map, all previously generated historical combustion simulation annotation data were adaptively calibrated, and the original annotation label content was updated to obtain a complete calibration annotation dataset. Based on the complete calibration and annotation dataset, the combustion simulation data annotation benchmark library of the digital twin system is updated synchronously, completing the linkage update of the digital twin iterative simulation process and the annotation data system.

2. The method as described in claim 1, characterized in that, The process involves analyzing and refining the combustion physics of dinitramide ammonium-based liquid propellants under different combustion environments and using historical annotation data from multiple batches. It also involves extracting the fixed correspondence between various combustion state descriptions and corresponding annotation labels during the combustion process, constructing an annotation association map containing combustion physics constraints for this type of propellant, and retaining the association records between combustion state descriptions and annotation labels, including: The combustion process observation records and multiple batches of historical labeled data of dinitramide ammonium-based liquid propellant under different oxygen contents, pressures and initial temperatures were collected. The observation records were then divided into three stages: combustion initiation, duration and termination, to obtain a set of segmented observation records arranged in chronological order. Each segment in the segmented observation record set is decomposed into elements, and three core contents are extracted: combustion morphology, medium reaction process, and product release. The extracted elements are then organized into a structured structured structured text set describing the physical laws of combustion. Each text entry in the structured description text set is matched one by one with the annotation tags of the multiple batches of historical annotation data. The correspondence between the combination of the three core contents and the annotation tags is mined to generate an initial correspondence list between the combustion state description items and the annotation tags. The list items are sorted according to the combustion process stage. Logical conflict checks are performed on the initial corresponding list. The labels corresponding to the same core content combination under different environments are compared. Duplicate or contradictory items are eliminated, and the order of the items is adjusted to be consistent with the combustion process stage, so as to obtain a fixed correspondence between the combustion state description item and the label. The fixed correspondence is grouped according to the environmental type, the combustion state description item is set as the starting content of the associated node, the label is set as the ending content, the connection relationship is defined according to the combustion physics law, and a labeled association map containing the combustion physics constraints of the dinitramide ammonium-based liquid propellant is generated. The starting node, ending node, and connection relationship of the labeled association map are serialized and stored to generate an association record between the combustion state description item and the label. The content of the association record is consistent with the nodes and relationships of the labeled association map.

3. The method as described in claim 2, characterized in that, The fixed correspondence is grouped according to environmental type, the combustion state description item is set as the starting content of the associated node, the label is set as the ending content, and the connection relationship is defined according to the combustion physics law to generate a labeled association map containing the combustion physics constraints of the dinitramide ammonium-based liquid propellant, including: Each combustion state description item in the fixed correspondence between the combustion state description item and the label is encoded in terms of content. The content encoding includes three types of identifiers: combustion process stage, environment type, and core element combination, generating a unique combustion state encoded text with consistent encoding length and format. Each of the label contents in the fixed correspondence is encoded. The content encoding includes three types of identifiers: label type, content category, and environment adaptation. A unique label encoded text is generated, and the encoding length is completely consistent with the combustion state encoded text. Each combustion state encoded text is bound to the corresponding label encoded text. The binding content includes information related to the environment type and the combustion process stage. A set of encoded binding pairs is generated, and the set of encoded binding pairs is sorted by both environment type and combustion process stage. The encoded binding pairs are grouped according to environment type, with the combustion state encoded text as the starting node and the labeled tag encoded text as the ending node, and the environment type as the connection attribute, to construct the initial framework of the labeled association graph. The nodes and attributes adopt a standardized data format. The initial framework of the labeled association map is logically optimized, the node connection relationship is adjusted to conform to the temporal logic of the combustion physics law, the node association paths of different combustion stages under the same environment are supplemented, and the optimized labeled association map is generated. The integrity of all nodes and connections in the optimized annotation association map is verified, isolated nodes or invalid connections are removed, and the annotation association map containing the combustion physical constraints of the dinitramide ammonium-based liquid propellant is generated.

4. The method as described in claim 1, characterized in that, The step of comparing the new combustion simulation data generated by the digital twin system through iterative simulation with the labeled association map, extracting combustion state description items that can be covered by the labeled association map and reusing the corresponding labels, and simultaneously separating new combustion state description items that are not matched by the labeled association map includes: The new combustion simulation data generated by the digital twin system is retrieved and segmented into three stages: combustion initiation, duration, and termination. Each segment contains information on the combustion morphology, reaction process, and product release of the corresponding stage. Multiple segments of combustion simulation sub-data are generated in the order of the process, and the segmentation boundaries are completely matched with the stage division of the combustion physics law. Each segment of the multi-segment combustion simulation sub-data is format-converted, converting the machine-readable format into a format that is completely consistent with the natural language description of the starting node of the labeled association map, and generating multi-segment standardized combustion description text. Each text entry in the multi-segment standardized combustion description text is compared one by one with all the starting nodes of the labeled association map. Each core element of the text entry is compared with the starting node, and a set of comparison results containing matching success or failure markers is generated. The markers are bound to the corresponding text entries one by one. Extract standardized combustion description text with a successful match marker from the comparison result set, associate it with the combustion state description item corresponding to the starting node, retrieve the corresponding annotation label content, and triple bind the combustion simulation sub-data, the combustion state description item, and the annotation label content to generate a result set of reusable annotation labels; The standardized combustion description text with matching failure markers is extracted from the comparison result set. The core elements of the text are extracted and the combination of combustion morphology, medium reaction process and product release is retained to generate a new combustion state description item that is not matched by the labeled association map. The result set of reused labels and the new combustion state description item are stored in a standardized manner. The storage path of the result set of reused labels is the same as the storage path of the multiple batches of historical labeled data. The new combustion state description item is stored in the path of the dataset to be labeled.

5. The method as described in claim 4, characterized in that, The step of binding the combustion simulation sub-data, the combustion state description item, and the label content together to generate a set of reusable label results includes: Retrieve standardized combustion description text with a successful match marker from the initial matching results of the reused label, associate it with the stage identifier and environment identifier of the corresponding combustion simulation sub-data, and generate a set of successfully matched text with associated attributes. The entries in the set of successfully matched text are sorted according to the combustion process stage. Each text entry in the successfully matched text set is matched with the starting node of the labeled association map to obtain the encoded text of the combustion state description item of the corresponding starting node. The text entry, the encoded text of the combustion state description item, the stage identifier, and the environment identifier are bound together to generate an encoded association set. The encoded association set is grouped according to the environment type. For the encoded text of each combustion state description item in the encoded association set, match the encoded text of the corresponding label in the label association map, retrieve the corresponding natural language label content, and bind the encoded text of the combustion state description item, the encoded text of the label, and the label content to generate a label association set; The annotation label content in the annotation label association set is bound to the corresponding combustion simulation sub-data. The binding content includes stage identifier, environment identifier, and combustion state description item, generating a combustion simulation sub-data set with reusable annotation labels. The set is sorted by both combustion process and environment type. The content of the combustion simulation sub-data set with reused labels is validated. The content of the labels is compared with the core information of the combustion simulation sub-data. Mismatched items are removed, and a result set of validated reused labels is generated. The result set of the verified reused label is stored in the labeled simulation data path of the digital twin system. The stored content includes the combustion simulation sub-data, the combustion state description item, the label content, the stage identifier, and the environment identifier.

6. The method as described in claim 4, characterized in that, The process involves extracting standardized combustion description text with matching failure markers from the comparison result set, refining the core elements of the text, retaining the combination of combustion morphology, medium reaction process, and product release, and generating new combustion state description items that are not matched by the labeled association map, including: The standardized combustion description text with matching failure markers is retrieved from the comparison result set, and associated with the stage identifier and environment identifier of the corresponding combustion simulation sub-data to generate a set of failed matching texts with associated attributes. The entries of the failed matching text set are sorted according to the combustion process stage. Each text entry in the failed matching text set is refined to extract three core contents: combustion mode, medium reaction process, and product release. Irrelevant or redundant information is removed to generate a core element combination set. Each entry in the core element combination set corresponds one-to-one with the failed matching text. Each combination entry in the core element combination set is compared one by one with the core element combination of all starting nodes in the labeled association map. If the combination entry does not appear in any starting node, an unmatched confirmation set is generated. The unmatched confirmation set entries include combination content, stage identifier, and environment identifier. Each combination entry in the unmatched confirmation set is given a structured description to generate an initial set of new combustion state description items; The initial set of the new combustion state description items is double-sorted according to the combustion process stage and the environmental type to generate an ordered set of new combustion state description items; The ordered set of new combustion state description items is stored in the unlabeled dataset path of the digital twin system. The stored content includes the new combustion state description items, stage identifiers, environmental identifiers, and sorting numbers.

7. The method as described in claim 1, characterized in that, The process involves targeted annotation of the separated new combustion state description items to obtain new annotation tags, integrating these new annotation tags with the structural logic of the associated records, and updating the combustion state description item-annotation tag associated records of the annotation association map, including: The ordered set of the new combustion state description items is retrieved, and the content of each new combustion state description item is parsed to clarify the combination relationship and environmental correlation attributes of combustion morphology, medium reaction process, product release, and generated parsed dataset to be labeled. The entries of the dataset to be labeled are arranged by sort number. Each entry in the dataset to be labeled is matched with the label content of similar combinations in the multiple batches of historical labeled data. Referring to the label framework structure of the multiple batches of historical labeled data and combining the combustion physics of the dinitramide ammonium-based liquid propellant, the labeling direction and content framework of each new combustion state description item are determined, and a labeling framework set is generated. According to each direction and frame of the annotation framework set, new annotation labels are generated for the corresponding new combustion state description item, so that the expression style and logical rules are consistent with the annotation labels of the multiple batches of historical annotation data, and a new annotation label set is generated. The items of the new annotation label set correspond one-to-one with the dataset to be annotated. The ordered set of new combustion state description items is bound to the corresponding entries of the newly added label set. The binding content includes stage identifier, environment identifier, and sort number, generating a set of binding pairs between the new combustion state description items and the newly added label. The set of binding pairs is arranged according to the sort number. The binding pairs of the new combustion state description item and the newly added label are validated. The correspondence between the core elements of the new combustion state description item and the content of the newly added label is compared. Mismatched binding pairs are eliminated, and a validated binding pair set is generated. The verified binding set is stored in the newly added temporary path of the digital twin system. The stored content includes the new combustion state description item, the newly added label, stage identifier, environment identifier, and sorting number.

8. The method as described in claim 7, characterized in that, The integration of the newly added annotation tags with the structural logic of the associated corresponding records includes: Retrieve the associated records of the combustion state description items and the label, parse the grouping rules, encoding methods and content organization order of the associated records, and generate a structural logic parsing report. The structural logic parsing report contains the complete organizational structure of the associated records. The content of each new combustion state description item in the set after verification is encoded. The encoding rules are completely consistent with the encoding rules of the combustion state description item. It includes three types of identifiers: combustion process stage, environment type, and core element combination, and generates a unique set of new combustion state encoded text. The content of each newly added annotation tag in the set after verification is encoded. The encoding rules are completely consistent with the encoding rules of the annotation tags, including three types of identifiers: annotation type, content category, and environment adaptation, to generate a unique set of encoded text for newly added annotation tags. The new combustion state encoded text set is bound to the corresponding entries of the newly added label encoded text set. The binding content includes stage identifier and environment identifier, generating a new set of encoded binding pairs. The new set of encoded binding pairs is grouped according to environment type and combustion process stage. According to the grouping rules of the associated corresponding records, the newly added code binding pair set is inserted into the corresponding group of the associated corresponding records, the order of the entries of the associated corresponding records is adjusted to be consistent with the combustion process stage, and an extended associated corresponding record containing the original and newly added correspondence is generated. The extended association corresponding records are validated by comparing the encoding rules, grouping rules, and content format of the new entries with those of the original entries, eliminating entries that do not conform to the rules, and generating validated extended association corresponding records. The update of the combustion state description item - label association corresponding record of the labeled association map includes: Retrieve the verified extended association corresponding record, convert the newly added code binding pair into a node and relationship format that the labeled association graph can recognize, including the starting node corresponding to the new combustion state code text, the ending node corresponding to the newly added label code text, and the connection relationship attribute corresponding to the environment type, and generate a graph update data set; Add the new start node and new end node from the updated map data set to the labeled association map, bind the environment type connection relationship attribute to the corresponding start and end nodes, and generate the initial version of the updated labeled association map. The updated annotation association graph initial version is subjected to logical consistency verification, which checks the encoding rules, grouping rules, and association logic of all nodes and connection relationships, removes nodes and relationships with encoding errors or logical contradictions, and generates a verified annotation association graph. All nodes and connections in the verified labeled association graph are serialized and stored according to the same storage rules as the corresponding association records, generating an updated initial version of the corresponding association records. By comparing the verified annotation association map with the updated association corresponding record initial version, entries with inconsistent content are removed, and the final updated association corresponding record of the combustion state description item and the annotation label is generated. The final updated combustion state description item and the associated record of the label are overwritten with the original associated record, and the verified label association map is overwritten with the original label association map. The data is then synchronously stored in the specified path of the digital twin system.

9. The method as described in claim 1, characterized in that, Based on the updated annotation association map, adaptive calibration is performed on all previously generated historical combustion simulation annotation data, including: Retrieve the updated annotation association map, extract the new start node, new end node, and new connection relationship attributes, and generate a new corresponding relationship set. The new corresponding relationship set is grouped according to environmental type and combustion process stage, and the grouping rules are consistent with the annotation association map. All previously generated historical annotation data of combustion simulation are retrieved and classified in two ways according to combustion process stage and environmental type to generate a multi-classified historical annotation dataset. The entries of the multi-classified historical annotation dataset are sorted by stage and environment, and the sorting logic is consistent with the annotation association map. Each entry in the multi-class historical annotation dataset is compared with the content of the newly added starting node in the newly added correspondence set one by one to identify the combustion state description item that perfectly matches the newly added starting node, and generate a set of historical annotation data entries with calibration marks. The set of historical labeled data entries with calibration marks is separated from the multi-class historical labeled dataset and stored separately from the entries without calibration marks to generate a historical labeled dataset to be calibrated and an uncalibrated historical labeled dataset. The storage paths of the two datasets are independent of each other. For each entry in the historical annotation dataset to be calibrated, the entry to be calibrated is bound to the new annotation label by matching the content of the new termination node of the new correspondence set with the content of the new starting node, i.e. the content of the new annotation label, to generate a calibration data set to be processed. The calibration data set to be processed is stored in the calibration temporary path of the digital twin system. The stored content includes the calibration entries, the newly added start node content, the newly added label, and the calibration mark. The storage format is consistent with the multiple batches of historical labeled data.

10. A data annotation system, characterized in that, The data annotation system includes a processor and a memory, wherein the memory stores a computer program, which is loaded and executed by the processor to implement the data annotation method for combustion simulation of dinitramide ammonium-based liquid propellants based on digital twins as described in any one of claims 1 to 9.

Images