Cross-domain heterogeneous data processing method for enterprise management building digital twin and related system

Abstract

This application relates to a method and related system for processing cross-domain heterogeneous data of a digital twin of an enterprise management building, including: acquiring a standard semantic library of the digital twin of the enterprise management building; receiving cross-domain heterogeneous process data from different management systems; acquiring system prompt words of a large language model; inputting the system prompt words and cross-domain heterogeneous process data into the large language model to obtain a business process model file output by the large language model; parsing the business process model file and extracting the node attributes of each process node from the business process model file; mapping the node attributes of each process node to the standard definitions and standard expressions corresponding to each management atom in the standard semantic library, so as to complete the semantic fusion of cross-domain heterogeneous process data from different management systems. The above method can automatically parse and transform cross-domain heterogeneous process data from different management systems into a unified standard format, thereby realizing the semantic fusion of cross-domain heterogeneous process data from different management systems.

Description

Technical Field

[0001] This application relates to the field of digital twin processing technology for building management, and in particular to a cross-domain heterogeneous data processing method and related system for enterprise building digital twins. Background Technology

[0002] In enterprise digital operations, building a digital twin of the management architecture that connects strategic planning and daily operations has become crucial for achieving panoramic business insights and intelligent decision-making. However, existing technologies face the following pressing technical challenges in achieving deep integration and dynamic mapping of cross-domain heterogeneous business processes: Different management systems, and even different versions of the same system, have significant differences in the definition and description of the same global business process, forming "semantic silos." Existing technologies lack the ability to standardize the semantic definition and unified description of global business processes, making it impossible to achieve the connection and integration of cross-domain heterogeneous process data at the semantic level. Summary of the Invention

[0003] Based on this, it is necessary to provide a cross-domain heterogeneous data processing method and related system for enterprise management building digital twins to address the above-mentioned technical problems. This method can automatically parse and transform cross-domain heterogeneous process data from different management systems into a unified standard format, thereby achieving semantic fusion of cross-domain heterogeneous process data from different management systems.

[0004] According to a first aspect of certain exemplary embodiments of this application, a method for processing cross-domain heterogeneous data of a digital twin of an enterprise management building is provided, comprising: obtaining a standard semantic library of the digital twin of the enterprise management building; the standard semantic library configuring standard definitions and standard expressions of multiple management atoms of the full-domain process of the enterprise management building; the management atoms being used to represent the node attributes of each process node in the business process of the full-domain process; the standard definitions being used to describe the semantic connotation of the management atoms; and the standard expressions being used to specify the unified expression form of the management atoms; receiving cross-domain heterogeneous process data from different management systems; obtaining system prompt words of a large language model; the system prompt words being used to instruct the large language model to output a business process model file containing multiple process nodes, and recording the node attributes of each process node in the business process model file; inputting the system prompt words and cross-domain heterogeneous process data into the large language model to obtain the business process model file output by the large language model; parsing the business process model file and extracting the node attributes of each process node from the business process model file; and mapping the node attributes of each process node to the standard definitions and standard expressions corresponding to each management atom in the standard semantic library, so as to complete the semantic fusion of cross-domain heterogeneous process data from different management systems.

[0005] Preferably, the node attributes of each process node are extracted from the business process model file, including: traversing the task nodes, gateway nodes, and event nodes in the business process model file; extracting the node attributes of each process node in the task nodes, gateway nodes, and event nodes respectively, whereby the node attributes include step name, business object, business metric, business rule, business risk, and business role; mapping the node attributes of each process node to the standard definitions and standard expressions corresponding to each management atom in the standard semantic library, including: mapping the step name of the task node to the standard definition of the management atom corresponding to the task node in the standard semantic library, and mapping the business object, business metric, business rule, business risk, and business role of the task node to the standard expression of the management atom corresponding to the task node; mapping the step name of the gateway node to the standard definition of the management atom corresponding to the gateway node in the standard semantic library, and mapping the business rule of the gateway node to the standard expression of the management atom corresponding to the gateway node; mapping the step name of the event node to the standard definition of the management atom corresponding to the event node in the standard semantic library, and mapping the business rule of the event node to the standard expression of the management atom corresponding to the event node.

[0006] Preferably, the system prompt is used to instruct the large language model to output a business process model file containing multiple process nodes in an XML structure according to the BPMN specification; traversing the task nodes, gateway nodes, and event nodes in the business process model file includes: traversing the task nodes, gateway nodes, and event nodes in the business process model file according to the XML structure of the business process model file and predefined node tags.

[0007] Preferably, the node attributes of each process node in the task node, gateway node, and event node are extracted respectively, including: extracting the step name of each process node from the inherent basic attributes of the task node, gateway node, and event node respectively; extracting predefined business object attribute fields, business indicator attribute fields, business rule attribute fields, business risk attribute fields, and business role attribute fields from the extended elements of the task node respectively, to obtain the business object, business indicator, business rule, business risk, and business role of the task node; when traversing to the gateway node, reading the element content of the expression element of the gateway judgment condition from the sequence stream of the gateway node, and using the read element content as the business rule of the gateway node; when traversing to the event node, parsing the trigger type and trigger content from the event definition sub-element of the event node, and constructing the business rule of the event node according to the trigger type and trigger content.

[0008] Preferably, when traversing an event node, the trigger type and trigger content are parsed from the event definition sub-element of the event node, and the business rules of the event node are constructed based on the trigger type and trigger content. This includes: when traversing to the start event of the event node, parsing the sub-element of the start event; if the trigger type of the start event is identified as a message trigger type based on the sub-element of the start event, then the message content is extracted from the trigger content of the sub-element of the start event, and the business rules of the start event are constructed based on the message trigger type and message content; if the trigger type of the start event is identified as a timed trigger type based on the sub-element of the start event, then the time period is extracted from the trigger content of the sub-element of the start event, and the business rules of the start event are constructed based on the timed trigger type and time period. If the child elements of an event identify the trigger type of the start event as a conditional trigger type, then the conditional expression is extracted from the trigger content of the child elements of the start event. The business rules for the start event are then constructed based on the conditional trigger type and the conditional expression. When traversing to the end event of an event node, the child elements of the end event are parsed. If the trigger type of the end event is identified as a termination type based on the child elements of the end event, then the process state is extracted from the trigger content of the child elements of the end event. The business rules for the end event are then constructed based on the termination type and the process state. If the trigger type of the end event is identified as a message output type based on the child elements of the end event, then the output message is extracted from the trigger content of the child elements of the end event. The business rules for the end event are then constructed based on the message output type and the output message.

[0009] Preferably, mapping the node attributes of each process node to the standard definitions and standard expressions corresponding to each management atom in the standard semantic library includes: semantically vectorizing the node attributes of each process node using a pre-trained domain word embedding model to obtain the semantic vectors of the node attributes of each process node; calculating the similarity between the semantic vectors of the node attributes of each process node and the semantic vectors in the semantic vector library of the standard semantic library, wherein each semantic vector in the semantic vector library corresponds to each management atom in the standard semantic library; taking the management atoms with similarity greater than the similarity threshold as the management atoms corresponding to the node attributes of each process node; and mapping the node attributes of each process node to the corresponding standard definitions and standard expressions of the management atoms.

[0010] Preferably, a cross-domain heterogeneous data processing method for a digital twin of an enterprise management building further includes: acquiring a historical mapping dataset, which contains multiple pairs of matching node attributes and management atoms through similarity calculation; calculating the distribution characteristic parameters of similarity based on the historical mapping dataset and through statistical analysis, where the distribution characteristic parameters include the mean and standard deviation; and determining a similarity threshold based on the mean and standard deviation.

[0011] According to a second aspect of certain exemplary embodiments of this application, a cross-domain heterogeneous data processing system for a digital twin of an enterprise management building is provided, comprising: a standard semantic library acquisition module, used to acquire a standard semantic library of the digital twin of the enterprise management building; the standard semantic library configures standard definitions and standard expressions of multiple management atoms of the full-domain process of the enterprise management building; management atoms are used to represent the node attributes of each process node in the business process of the full-domain process; standard definitions are used to describe the semantic connotation of management atoms; and standard expressions are used to specify the unified expression form of management atoms; a receiving module, used to receive cross-domain heterogeneous process data from different management systems; and a system prompt word acquisition module, used to acquire system prompts from a large language model. The system prompts are used to instruct the large language model to output a business process model file containing multiple process nodes, and to record the node attributes of each process node in the business process model file. The business process model file acquisition module is used to input the system prompts and cross-domain heterogeneous process data into the large language model to obtain the business process model file output by the large language model. The node attribute extraction module is used to parse the business process model file and extract the node attributes of each process node from the business process model file. The data mapping module is used to map the node attributes of each process node to the standard definitions and standard expressions corresponding to each management atom in the standard semantic library, so as to complete the semantic fusion of cross-domain heterogeneous process data from different management systems.

[0012] According to a third aspect of certain exemplary embodiments of this application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any of the above methods.

[0013] According to a fourth aspect of certain exemplary embodiments of this application, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0014] The aforementioned method and system for cross-domain heterogeneous data processing of a digital twin of an enterprise management building includes: acquiring a standard semantic library for the digital twin of the enterprise management building; configuring standard definitions and standard expressions for multiple management atoms of the full-domain process of the enterprise management building in the standard semantic library; using management atoms to represent the node attributes of each process node in the business process of the full-domain process; using standard definitions to describe the semantic connotation of management atoms; and using standard expressions to specify the unified expression form of management atoms; receiving cross-domain heterogeneous process data from different management systems; acquiring system prompt words from a large language model; using system prompt words to instruct the large language model to output a business process model file containing multiple process nodes, and recording the node attributes of each process node in the business process model file; inputting the system prompt words and cross-domain heterogeneous process data into the large language model to obtain the business process model file output by the large language model; parsing the business process model file and extracting the node attributes of each process node from the business process model file; and mapping the node attributes of each process node to the standard definitions and standard expressions corresponding to each management atom in the standard semantic library to complete the semantic fusion of cross-domain heterogeneous process data from different management systems. Therefore, by leveraging the standard semantic library of the enterprise management building's digital twin, the process elements of the enterprise management building's overall processes are broken down into fine-grained management atoms, improving the matching accuracy of cross-domain heterogeneous process data from different management systems. Furthermore, the large language model can automatically parse and transform cross-domain heterogeneous process data from different management systems into a unified standard format business process model file, thereby achieving semantic fusion of cross-domain heterogeneous process data from different management systems through data mapping. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating a cross-domain heterogeneous data processing method for a digital twin of an enterprise management building, as shown in some exemplary embodiments of this application. Figure 2 This is a flowchart illustrating a cross-domain heterogeneous data processing method for a digital twin of an enterprise management building, as shown in some other exemplary embodiments of this application. Figure 3 A network panorama diagram illustrating a cross-domain heterogeneous data processing method for a digital twin of an enterprise management building, as shown in some exemplary embodiments of this application; Figure 4 This is a structural block diagram of a cross-domain heterogeneous data processing system for a digital twin of an enterprise management building, as shown in some other exemplary embodiments of this application. Figure 5 This is a diagram of the internal structure of an electronic device in some other exemplary embodiments of this application. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0017] The following detailed descriptions are provided to aid the reader in gaining a comprehensive understanding of the methods, apparatus, electronic devices, storage media, and / or computer program products described herein. However, after understanding the disclosure of this application, various changes, modifications, and equivalents of the methods, apparatus, storage media, and / or computer program products described herein will become apparent. For example, the order of operations described herein is merely illustrative and is not limited to those orders set forth herein, but may be changed as will become clear after understanding the disclosure of this application, except for operations that must occur in a specific order. Furthermore, for clarity and conciseness, descriptions of features known in the art may be omitted.

[0018] The features described herein may be implemented in different forms and should not be construed as limited to the examples described herein. Rather, the examples described herein are provided only to illustrate some of the many feasible ways of implementing the methods, electronic devices, and / or storage media described herein, many of which will become clear upon understanding this application.

[0019] The terminology used herein is for the purpose of describing various examples only and is not intended to limit disclosure. Unless the context clearly indicates otherwise, the singular form is intended to include the plural form as well. The terms “comprising,” “including,” and “having” indicate the presence of the described features, quantities, operations, components, elements, and / or combinations thereof, but do not exclude the presence or addition of one or more other features, quantities, operations, components, elements, and / or combinations thereof. Unless otherwise stated, “ / ” means “or,” for example, A / B can mean A or B; “and / or” in the text is merely a description of the relationship between related objects, indicating that three relationships can exist, for example, A and / or B can mean: A alone, A and B simultaneously, and B alone. Furthermore, in the description of embodiments of the invention, “multiple” means two or more.

[0020] Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains upon understanding this application. Unless expressly defined herein, terms (such as those defined in a general dictionary) shall be interpreted as having a meaning consistent with their meaning in the context of the relevant field and in this application, and shall not be interpreted in an idealized or overly formalistic manner.

[0021] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in some of the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0022] Furthermore, in the description of the examples, detailed descriptions of well-known related structures or functions will be omitted when it is believed that such detailed descriptions would lead to a vague interpretation of this application.

[0023] In the following description, embodiments will be described in detail with reference to the accompanying drawings. However, embodiments may be implemented in various forms and are not limited to the examples described herein.

[0024] In some exemplary embodiments of this application, a method for cross-domain heterogeneous data processing of a digital twin for enterprise building management is provided. For example... Figure 1 The method for cross-domain heterogeneous data processing of a digital twin for enterprise management buildings, as shown, includes the following steps: Step S101: Obtain the standard semantic library of the enterprise management building's digital twin. The standard semantic library configures the standard definitions and standard expressions of multiple management atoms in the enterprise management building's full-domain process. Management atoms are used to represent the node attributes of each process node in the business process in the full-domain process. Standard definitions are used to describe the semantic connotation of management atoms, and standard expressions are used to specify the unified expression form of management atoms.

[0025] In this embodiment, a standard semantic library for the digital twin of the enterprise management building is first constructed. The standard semantic library configures the standard definitions and standard expressions of multiple management atoms of the enterprise management building's full-domain processes.

[0026] Specifically, a top-down domain analysis approach is adopted to construct a standard semantic library for the entire domain process of the enterprise management architecture's digital twin. Building this standard semantic library is a prerequisite for performing cross-domain process data semantic fusion and association mapping. The standard semantic library covers the management atoms of the entire domain process across the six major modules of the enterprise management architecture. The six major modules include strategy, management system, enterprise architecture, enterprise value stream, organizational structure, and operational processes. Among them, strategy includes a strategy map, strategic theme target cards, strategic packages, and annual strategic implementation plan information. The management system includes a management system framework, basic management system information, module information, management elements, management capabilities, a six-integration matrix, systems, and standards. Enterprise architecture includes business architecture (business rules, business objects, business metrics, business roles, business steps, business levels, business capabilities, operational-level business processes, professional domains, and a list of business capability enablements), application architecture (application modules, application groups, application domains, and application function items / sub-items), data architecture (data domains, data themes, conceptual entities, logical entities, and attributes), technology architecture (technology domains, technology levels, technology services, and technology components), and security architecture (security domains, security levels, security services, and security components). Enterprise value streams include enterprise-level value streams, professional-level process groups, and end-to-end processes. Organizational structure includes departments and organizational positions. Designed and operational processes include process definitions, process instances, task instances, operators, and business forms.

[0027] The standard semantic library consists of multiple management atoms. Management atoms include process step names, business objects, business metrics, business rules, business risks, and business roles. The standard semantic library defines standard definitions and expressions for each type of management atom to ensure that process data across systems and states can be mapped to a unified semantic benchmark.

[0028] Among them, the standard semantic library covers only the business processes involving the entire domain process. Each process node in each module of the six major modules of the enterprise management building can be broken down into several management atoms.

[0029] Take the emergency repair work order business process as an example. This process belongs to the end-to-end process under the enterprise value stream module, including process nodes such as whether the estimated repair time exceeds 12 hours, initiating repair and filling out the work order, granting permission, transferring to the corresponding work order, repair completion, and work order archiving. These process nodes can be further broken down into multiple management atoms: Step names: Initiate emergency repair, fill out work order, approve work permit, execute emergency repair, file work order; Business targets: Emergency repair work orders, faulty equipment; Business performance indicators: Repair time ≤ 12 hours, response time ≤ 30 minutes; Business rule: If the estimated repair time exceeds 12 hours, it must be converted to a regular work order process; Business risks: Risk of failure to complete within the time limit; risk of execution without approval; Business roles: emergency repair personnel, dispatchers, and approvers.

[0030] After the above decomposition, the process nodes can be transformed into a set of management atoms with a unified structure, and then semantically aligned with the management atoms in the standard semantic library.

[0031] Step S102: Receive cross-domain heterogeneous process data from different management systems.

[0032] In this embodiment, heterogeneous process data is received from different management systems or different state processes within the same management system. Heterogeneous process data includes structured data, semi-structured data, and unstructured data. Structured data includes forms, semi-structured data includes log data, and unstructured data includes text and images.

[0033] Step S103: Obtain the system prompt words of the large language model. The system prompt words are used to instruct the large language model to output a business process model file containing multiple process nodes, and record the node attributes of each process node in the business process model file.

[0034] In this embodiment, the cross-modal reasoning capabilities of large language models, such as image recognition and semantic understanding, are utilized to automatically parse the implicit process logic and business context in cross-domain heterogeneous process data from different management systems, generating a standard business process model file in one go. The business process model file can be a BPMN file.

[0035] Regarding the selection of the large language model, a general-purpose large language model that supports multimodal input is adopted, such as the DeepSeek series models, the Qwen series models, or models with equivalent capabilities. The large language model possesses image recognition capabilities for processing flowchart images. It also has long text understanding capabilities for processing unstructured text such as user manuals. Furthermore, it has structured output capabilities for generating BPMN (Business Process Model and Notation) files in XML (eXtensible Markup Language) format.

[0036] For large language models, system prompts need to be pre-configured to specify the XML structure format, field names of node attributes, and filling specifications of the output of the large language model.

[0037] Step S104: Input the system prompt words and cross-domain heterogeneous process data into the large language model to obtain the business process model file output by the large language model.

[0038] In this embodiment, system prompts and cross-domain heterogeneous process data are input into the large language model. The large language model can process the cross-domain heterogeneous process data according to the system prompts, output a business process model file with multiple process nodes, and record the node attributes of each process node in the business process model file.

[0039] Specifically, cross-domain heterogeneous process data is uniformly encapsulated into an input format acceptable to the large language model. Then, preset system prompts are appended, instructing the large language model to output an XML-structured business process model file according to the BPMN 2.0 specification, and at each process node... <extensionelements>The extended fields record node attributes such as the business object, key indicators, execution rules, potential risks, and collaborating personnel requirements corresponding to each process node. Finally, the large language model returns a complete BPMN file in one go, which the system then validates for format before storing it in the dataset.

[0040] Business process model files, or BPMN files, not only fully define the names and flow relationships of each process node, but also record other node attributes in each process node, such as business objects, key indicators, execution rules, potential risks and collaborating personnel requirements, thereby realizing the unified construction of a standard process model rich in business semantics from multi-source heterogeneous raw data.

[0041] Step S105: Parse the business process model file and extract the node attributes of each process node from the business process model file.

[0042] In this embodiment, the business process model file is parsed, such as the XML structure of a BPMN file, and all process nodes of the business process model file are traversed from the XML structure. For each process node, its node attributes are automatically extracted.

[0043] The system dynamically categorizes extracted node attributes into corresponding management atom sets based on preset classification mapping rules, such as a semantic tagging system based on business ontology. For example, management atom sets include step names, business objects, business metrics, business rules, business risks, and business roles. This classification result is stored in the cross-domain process input dataset as the basis for subsequent association and mapping with the standard semantic library.

[0044] Step S106: Map the node attributes of each process node to the standard definitions and standard expressions corresponding to each management atom in the standard semantic library to complete the semantic fusion of cross-domain heterogeneous process data from different management systems.

[0045] In this embodiment, each management atom in the standard semantic library is used to represent the node attributes of each process node in the business process across the entire domain. The node attributes of each process node are mapped to the standard definitions and standard expressions of the management atoms corresponding to each process node in the standard semantic library, thereby ensuring that process data across systems and states can be mapped to a unified semantic benchmark, and completing the semantic fusion of cross-domain heterogeneous process data from different management systems.

[0046] The aforementioned method for processing cross-domain heterogeneous data in enterprise management building digital twins utilizes a standard semantic library for the enterprise management building digital twin to decompose the process elements of the entire domain process of the enterprise management building into fine-grained management atoms, thereby improving the matching accuracy of cross-domain heterogeneous process data from different management systems. Furthermore, a large language model can automatically parse and transform cross-domain heterogeneous process data from different management systems into a unified standard format business process model file, and then achieve semantic fusion of cross-domain heterogeneous process data from different management systems through data mapping.

[0047] In one embodiment, step S105 above, extracting node attributes of each process node from the business process model file, includes: traversing task nodes, gateway nodes, and event nodes in the business process model file; extracting node attributes of each process node in the task nodes, gateway nodes, and event nodes respectively. Node attributes include step name, business object, business metric, business rule, business risk, and business role. Correspondingly, step S106 above, mapping the node attributes of each process node to the standard definitions and standard expressions corresponding to each management atom in the standard semantic library, includes: mapping the step name of the task node to the standard definition of the management atom corresponding to the task node in the standard semantic library, and mapping the business object, business metric, business rule, business risk, and business role of the task node to the standard expression of the management atom corresponding to the task node; mapping the step name of the gateway node to the standard definition of the management atom corresponding to the gateway node in the standard semantic library, and mapping the business rule of the gateway node to the standard expression of the management atom corresponding to the gateway node; mapping the step name of the event node to the standard definition of the management atom corresponding to the event node in the standard semantic library, and mapping the business rule of the event node to the standard expression of the management atom corresponding to the event node.

[0048] Specifically, task nodes ( <bpmn:task>Task nodes correspond to the actual business actions that personnel need to perform in the business process. Task nodes are the main source of management atomic extraction. Taking the emergency repair work order process as an example, initiating the emergency repair and filling in the work order, work permit approval, emergency repair execution, and work order archiving are all task nodes.

[0049] Gateway node ( <bpmn:exclusivegateway>This corresponds to conditional judgments and branch selections in the business process. Gateway nodes mainly correspond to the management atoms of business rule classes; their judgment conditions are the rule content, and the branching path constitutes the rule's execution action. Taking the emergency repair work order process as an example, the gateway node is the exclusive gateway node for determining whether the estimated repair time exceeds 12 hours. The subsequent path is determined based on the judgment result: if the repair time does not exceed 12 hours, the emergency work order process continues; otherwise, it switches to the regular work order process.

[0050] Event nodes primarily correspond to the triggering and termination conditions in business rules. Event nodes are divided into start events (...). <bpmn:startevent>) and end event ( <bpmn:endevent>The start event corresponds to the triggering condition in the business process. Taking the emergency repair work order process as an example: receiving a fault report is the start event for this process node. The end event corresponds to the termination state in the business process. Taking the emergency repair work order process as an example: the work order archiving is completed is the end event for this process node.

[0051] The node attributes of each process node are mapped to the standard definitions and expressions corresponding to each management atom in the standard semantic library. Taking the management atoms in the "Emergency Repair Work Order Process" as an example, the standard definitions and expressions of various management atoms are explained as follows: Management atoms corresponding to business metrics. Standard definition: The maximum allowed time from task triggering to completion. Standard expression: "Processing time limit ≤ N minutes" (N is a specific value). For example, different expressions such as "completed within two hours" or "no more than 120 minutes" appearing in business process data are all standardized to "processing time limit ≤ 120 minutes" after mapping.

[0052] The management atom corresponding to the business rule. Standard definition: The precondition for triggering a process branch. Standard expression: When the [condition] is met, execute the [action]. For example, expressions such as "transfer to regular ticket if it exceeds 12 hours" and "change ticket if the time is long" in the business process data are uniformly standardized after mapping to "when the emergency repair time is > 12 hours, transfer to regular work ticket process".

[0053] Management atoms corresponding to business roles. Standard definition: The organizational unit responsible for executing this step. Standard expression: [Department] + [Job level / role]. Descriptions of emergency repair teams, on-site operators, etc., in business process data are standardized as "Operations and Maintenance Department + Emergency Repair Personnel" after mapping.

[0054] Management atoms corresponding to business risks. Standard definition: Potential factors that may lead to deviations or losses during process execution. Standard expression: [Risk type]: [Triggering condition] → [Consequence]. For example, vague expressions such as "failure to complete within time limit" in business process data are standardized after mapping to "Compliance risk: Execution without approval → Safety liability incident".

[0055] Through the above standard definitions and expressions, regardless of which management system the original business process data comes from or what format it is described in, it can be mapped to the corresponding unified expression in the standard semantic library, thereby realizing the semantic fusion and association mapping of cross-domain heterogeneous process data.

[0056] In one embodiment, a system prompt is used to instruct the large language model to output a business process model file containing multiple process nodes in an XML structure, according to the BPMN specification. The steps described above for traversing the task nodes, gateway nodes, and event nodes in the business process model file include: traversing the task nodes, gateway nodes, and event nodes in the business process model file based on the XML structure and predefined node tags.

[0057] In this embodiment, a system prompt instructs the large language model to output a BPMN file with a unified XML structure. Each process node in the BPMN file has predefined node tags. By traversing the XML structure and predefined node tags of the BPMN file, the task nodes, gateway nodes, and event nodes in the business process model file are extracted.

[0058] Specifically, the system identifies node types by parsing the structure and predefined tags of the BPMN 2.0 standard XML file. For example: Task Node: By identifying elements in the XML structure <bpmn:task>The label is used for location. This label clearly identifies a task or activity that needs to be performed.

[0059] Gateway nodes: located by recognizing various gateway tags in the XML structure. For example... <bpmn:exclusivegateway>Indicates an exclusive gateway. <bpmn:parallelgateway>These tags represent parallel gateways, etc. They define the decisions, branches, and convergence points in the process.

[0060] Event nodes: located by identifying event tags within the XML structure. For example... <bpmn:startevent>Indicates the start of the event. <bpmn:endevent>These labels indicate events such as termination. They represent the triggering, termination, or intermediate state changes of a process.

[0061] In one embodiment, the steps of extracting the node attributes of each process node in the task node, gateway node, and event node respectively include: extracting the step name of each process node from the inherent basic attributes of the task node, gateway node, and event node respectively; extracting predefined business object attribute fields, business indicator attribute fields, business rule attribute fields, business risk attribute fields, and business role attribute fields from the extended elements of the task node to obtain the business object, business indicator, business rule, business risk, and business role of the task node; when traversing to the gateway node, reading the element content of the expression element of the gateway judgment condition from the sequence stream of the gateway node, and using the read element content as the business rule of the gateway node; when traversing to the event node, parsing the trigger type and trigger content from the event definition sub-element of the event node, and constructing the business rule of the event node based on the trigger type and trigger content.

[0062] Specifically, node attribute extraction is based on the standard structure of BPMN XML and custom extended fields. Inherent basic attributes are extracted: for any type of process node, its inherent basic attributes are extracted. These inherent basic attributes include the process node's name attribute, which is directly used as the "step name" of that process node.

[0063] Extract extended attributes: The system extracts these attributes from the task node. <extensionelements>Within the extended element, predefined custom business attribute fields are extracted. For example, the business object attribute field, business metric attribute field, business rule attribute field, business risk attribute field, and business role attribute field correspond to the business object, business metric, business rule, business risk, and business role of the task node, respectively.

[0064] For gateway nodes, their core decision conditions are typically stored in expression elements. The system reads the content of the expression elements of the gateway node's decision conditions from the sequence stream and uses the read content as the gateway node's business rules.

[0065] For an event node, its event definition sub-element needs to be parsed. The trigger type and trigger content are then parsed from the event definition sub-element of the event node, and the business rules of the event node are constructed based on the trigger type and trigger content.

[0066] Furthermore, the steps described above, when traversing to an event node, parsing the trigger type and trigger content from the event definition sub-elements of the event node, and constructing the business rules of the event node based on the trigger type and trigger content, include: when traversing to the start event of the event node, parsing the sub-elements of the start event; if the trigger type of the start event is identified as a message trigger type based on the sub-elements of the start event, then extracting the message content from the trigger content of the sub-elements of the start event, and constructing the business rules of the start event based on the message trigger type and message content; if the trigger type of the start event is identified as a timed trigger type based on the sub-elements of the start event, then extracting the time period from the trigger content of the sub-elements of the start event, and constructing the business rules of the start event based on the timed trigger type and time period; if... If the child elements of the start event identify the trigger type as a conditional trigger type, then the conditional expression is extracted from the trigger content of the child elements of the start event. The business rules for the start event are then constructed based on the conditional trigger type and the conditional expression. When traversing to the end event of the event node, the child elements of the end event are parsed. If the trigger type of the end event is identified as a termination type based on the child elements of the end event, then the process state is extracted from the trigger content of the child elements of the end event. The business rules for the end event are then constructed based on the termination type and the process state. If the trigger type of the end event is identified as a message output type based on the child elements of the end event, then the output message is extracted from the trigger content of the child elements of the end event. The business rules for the end event are then constructed based on the message output type and the output message.

[0067] Specifically, for a start event: Parse the child elements of the start event. Identify the trigger type of the start event based on the child elements. Trigger types include message trigger type, timed trigger type, and conditional trigger type. Further, extract the trigger content from the child elements of the start event. Trigger content includes message content, time period, and conditional expression. Construct the business rules for the start event based on the message trigger type and message content. And / or, construct the business rules for the start event based on the timed trigger type and time period. And / or, construct the business rules for the start event based on the conditional trigger type and conditional expression. For example, a start event triggered by a message might have the following rule extracted: "When a 'device fault alarm' message is received, the process starts." For termination events: Parse the child elements of the termination event. Identify the trigger type of the termination event based on the child elements. Trigger types include termination type and message output type. Trigger content includes process status and output message. Construct business rules for the termination event based on the termination type and process status. And / or, construct business rules for the termination event based on the message output type and output message. For example, for a termination event with an output message, the rule can be extracted as: "When the process is completed, send a 'work order closed' notification to the 'Operations and Maintenance Management Platform'".

[0068] In one embodiment, step S106, mapping the node attributes of each process node to the standard definitions and standard expressions corresponding to each management atom in the standard semantic library, includes: performing semantic vectorization processing on the node attributes of each process node using a pre-trained domain word embedding model to obtain the semantic vectors of the node attributes of each process node; calculating the similarity between the semantic vectors of the node attributes of each process node and the semantic vectors in the semantic vector library of the standard semantic library, wherein each semantic vector in the semantic vector library corresponds to each management atom in the standard semantic library; taking the management atoms with similarity greater than the similarity threshold as the management atoms corresponding to the node attributes of each process node; and mapping the node attributes of each process node to the corresponding standard definitions and standard expressions of the management atoms.

[0069] The steps for determining the similarity threshold include: obtaining a historical mapping dataset, which contains multiple pairs of matching node attributes and management atoms calculated through similarity; calculating the distribution characteristic parameters of similarity based on the historical mapping dataset and through statistical analysis, including the mean and standard deviation; and determining the similarity threshold based on the mean and standard deviation.

[0070] Specifically, the previously categorized management atom sets are mapped to management atoms in the standard semantic library. The management atom sets are traversed, and the attributes of each node in each set are semantically vectorized. This is achieved by generating high-dimensional semantic vectors using a pre-trained domain word embedding model (such as Sentence-BERT). Based on a predefined semantic vector library in the standard semantic library, a cosine similarity algorithm is used to calculate the similarity score between the semantic vector of the current node's attribute and the vectors of each entry in the semantic vector library. Assuming a similarity threshold of 0.85, when the similarity score exceeds the threshold, the current node's attribute is mapped to the corresponding standardized expression in the standard semantic library. For example, "within two hours" and "no more than 120 minutes" are normalized to the unified "processing time limit ≤ 120 minutes" in the standard semantic library. "Power outage repair" and "equipment repair" are uniformly mapped to "fault repair".

[0071] Furthermore, based on the historical mapping dataset (containing verified matching pairs of node attributes and management atoms), the distributional characteristic parameters of cosine similarity are calculated through statistical analysis. These distributional characteristic parameters include the mean μ and standard deviation σ, dynamically generating a similarity threshold. The similarity threshold T is calculated as T = μ + kσ. The optimization coefficient k is iteratively solved using a binary search algorithm within a preset interval [0.1, 0.5] to maximize matching accuracy and minimize the false matching rate; that is, iteration stops when k results in an accuracy ≥ 95% and a false matching rate ≤ 5%. After each node attribute and management atom association mapping, the system automatically incorporates the current matching result into the historical mapping dataset, triggering a closed-loop update mechanism for the relevant parameters of the similarity threshold. This achieves dynamic adaptive optimization of the similarity threshold, ensuring that the confidence parameter for cosine similarity calculation is always based on a real-time optimized statistical model.

[0072] The following explanation uses the atomic mapping process of emergency repair work order management as an example: First, the initial similarity threshold is calculated: Assume the system has accumulated historical matching data for 500 manually verified management atoms. Statistical analysis is performed on the cosine similarity values ​​of these 500 matching data pairs, calculating the mean μ, assumed to be 0.87. The standard deviation σ is 0.05. Taking the initial optimization coefficient k=0.3, the initial dynamic similarity threshold T=0.87+0.3×0.05=0.885.

[0073] Then the binary search iterative process begins: the system backtests the historical mapping dataset with a T value of 0.885 to obtain the matching accuracy and mismatch rate. Assuming that the stopping condition of accuracy ≥ 95% and mismatch rate ≤ 5% has not yet been met, the iteration begins.

[0074] First iteration: k is set to the midpoint of the interval [0.3, 0.5] at 0.4, T = 0.87 + 0.4 × 0.05 = 0.890. The backtest results show that the accuracy still does not meet the conditions, so the iteration continues.

[0075] Second iteration: k is set to the midpoint of the interval [0.4, 0.5] at 0.45, T = 0.87 + 0.45 × 0.05 = 0.8925. The backtesting results show that the accuracy meets the stopping condition, and the iteration ends. Finally, the dynamic similarity threshold is determined.

[0076] Finally, the current batch is mapped and updated in real time: using T=0.8925 as the similarity threshold, the node attributes extracted from the emergency repair work order process are mapped to management atoms. Taking the overdue risk in the business risk set as an example: the system calculates that the cosine similarity between its semantic vector and "timeliness risk: task not completed within the specified time limit" in the standard semantic library is higher than the similarity threshold, so the mapping is successful. The standardized result in the standard semantic library is "timeliness risk: repair time exceeds the threshold → process escalation". For the business role "dispatcher", the calculated cosine similarity with "operations and maintenance dispatcher" in the standard semantic library is higher than the similarity threshold, so the mapping is successful. The standardized result in the standard semantic library is "operations and maintenance department + dispatcher".

[0077] After this batch of mapping is completed, the system automatically incorporates the newly added matching results into the historical mapping dataset, expanding the size of the historical mapping dataset from 500 pairs to the corresponding number. The system then triggers a closed-loop update of the similarity threshold parameters: recalculating the mean μ and standard deviation σ of the expanded historical mapping dataset, and re-performing the binary search with the updated parameters to obtain a new dynamic similarity threshold T for use in the next batch of mapping. This iterative process ensures that the similarity threshold is always dynamically calibrated based on the latest historical data, continuously improving the threshold accuracy.

[0078] Furthermore, large language models can be used to automate secondary verification. Periodically, 5% of the standardized management atoms are randomly sampled using a stratified sampling strategy. The large language model API is then used for deep semantic comparison, requiring the model to output structured matching probabilities and conflict annotations. Based on the matching probability distribution output by the large language model, a confidence level ≥ 0.95 is considered valid, and the system calculates the average matching accuracy across the entire process. If the accuracy is ≥ 95%, the process standardization is confirmed; otherwise, parameter optimization is automatically performed until the quality threshold is met.

[0079] Furthermore, during the cosine similarity matching of extracted node attributes with management atoms in the standard semantic library, if the matching score is still lower than a preset threshold after adaptive optimization and automated verification (e.g., the preset threshold is 0.5), the node attribute is automatically identified according to a predefined set of management atoms and structurally stored in the "Exception Management" library. Simultaneously, the node attribute is temporarily retained in the standardized process. This is periodically reviewed manually, and the node attributes in the "Exception Management" library are pushed to the review terminal. Reviewers conduct semantic rationality assessments based on the standard semantic library and domain expertise. If the semantics are deemed valid and the process data does not previously exist, the standard semantic library is dynamically updated after specifying a standardized expression.

[0080] In one embodiment, after step S106 above, a cross-domain heterogeneous data processing method for enterprise management building digital twins further includes: serializing each management atom in the standard semantic library according to the logical topology structure to generate a standardized process file based on the predefined node flow relationship in the business process model file; wherein, the node flow relationship includes the sequential flow identifiers of the start node, end node, gateway branch, and parallel path.

[0081] Specifically, the business process model file is a BPMN file, and the standardized process file is a standardized BPMN process file. Based on the predefined node flow relationships in the BPMN file (including the sequence flow identifiers of start nodes, end nodes, gateway branches, and parallel paths), the mapped standardized management atoms are re-serialized according to the logical topology and concatenated to generate a standardized BPMN process file that fully conforms to the standard semantic library specification. Furthermore, the standardized process data files mapped to the standard semantic library are categorized and stored in different systems.

[0082] Regarding the cross-domain heterogeneous data processing method for a digital twin of an enterprise management building as described in the above embodiments, a specific embodiment is given below. For example... Figure 2 As shown, a method for cross-domain heterogeneous data processing of a digital twin for enterprise management buildings includes the following steps: Step S201: Establish a standard semantic library for managing the entire building area.

[0083] Step S202: Cross-domain heterogeneous process data conversion.

[0084] Specifically, step S202 includes: Step S2021: Call the large language model interface; Step S2022: Create standardized process documents.

[0085] Step S203: Extract and classify the node attributes corresponding to the management atoms in the cross-domain heterogeneous process data.

[0086] Step S204: Association and mapping of node attributes with management atoms in the standard semantic library.

[0087] Specifically, step S204 includes: Step S2041, semantic vectorization of node attributes; Step S2042, cosine similarity matching and mapping.

[0088] Step S205: Adaptive optimization and verification of association mapping.

[0089] Step S206: Special case storage and exception management of node attributes.

[0090] Step S207: Automated secondary verification of the large language model.

[0091] Step S208: Standardized process reconstruction and storage.

[0092] One example is a panoramic view of a cross-domain heterogeneous data processing method for enterprise management building digital twins, as shown below. Figure 3 As shown.

[0093] In summary, this application overcomes the limitations of existing technologies in handling only single data structures in cross-domain process processing, achieving unified parsing and fusion of multimodal heterogeneous process data. Leveraging the cross-modal semantic understanding and reasoning capabilities of a large language model, it encapsulates and jointly parses multi-source heterogeneous data, including structured and semi-structured data, in a single process. This technology automatically extracts and integrates the process logic and attributes implicit in unstructured data, such as node names, key indicators, and execution rules, transforming them into standard process files rich in business semantics. This not only improves the completeness and accuracy of process modeling but also makes unstructured data such as images and free text, which were previously unrecognizable by traditional methods, effective inputs for process association mapping, truly achieving seamless transformation of multi-source heterogeneous data into a unified standard process model.

[0094] This application constructs a standardized mapping system for "management atoms," transforming scattered professional fields into standardized semantic units with unified definitions. This solves the problem that existing technologies cannot eliminate cross-professional naming ambiguities, enabling direct semantic alignment of data between systems and avoiding "information silos" and logical breaks caused by inconsistent naming.

[0095] This application establishes a closed-loop mechanism of "standard library construction - semantic mapping - operational verification," effectively solving the challenge of topological relationships (standards, design, operation) between different management systems and between different state processes within the same system in modern enterprise management. This process not only eliminates the semantic gap caused by system differences and state variations but also integrates previously fragmented multi-source data into unified, machine-recognizable standardized process documents. This achieves end-to-end semantic connectivity from macro-strategic blueprints to micro-operational data, ensuring that the digital twin can perform panoramic perception and dynamic simulation of cross-domain business based on a unified standard language.

[0096] It should be understood that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order constraint on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0097] In some exemplary embodiments of this application, such as Figure 4 As shown, a cross-domain heterogeneous data processing system for an enterprise management building digital twin is provided, including a standard semantic library acquisition module 401, a receiving module 402, a system prompt word acquisition module 403, a business process model file acquisition module 404, a node attribute extraction module 405, and a data mapping module 406. The standard semantic library acquisition module 401 is used to acquire the standard semantic library of the enterprise management building digital twin. The standard semantic library configures the standard definitions and standard expressions of multiple management atoms in the full-domain process of the enterprise management building. Management atoms are used to represent the node attributes of each process node in the business process of the full-domain process. Standard definitions are used to describe the semantic connotation of management atoms, and standard expressions are used to specify the unified expression form of management atoms. The receiving module 402 is used to receive cross-domain heterogeneous process data from different management systems. The system prompt word acquisition module 403 is used to acquire the system prompt words of the large language model. The system prompt words are used to indicate that the output of the large language model contains multiple... The system includes a business process model file for each process node, which records the node attributes of each process node within the business process model file. A business process model file acquisition module 404 is used to input system prompts and cross-domain heterogeneous process data into a large language model to obtain the business process model file output by the large language model. A node attribute extraction module 405 is used to parse the business process model file and extract the node attributes of each process node from it. A data mapping module 406 is used to map the node attributes of each process node to the standard definitions and expressions corresponding to each management atom in the standard semantic library, thereby completing the semantic fusion of cross-domain heterogeneous process data from different management systems.

[0098] Specific limitations regarding the cross-domain heterogeneous data processing system for a digital twin of an enterprise management building can be found in the limitations of the cross-domain heterogeneous data processing method for a digital twin of an enterprise management building described above, and will not be repeated here. Each module in the aforementioned cross-domain heterogeneous data processing system for a digital twin of an enterprise management building can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in an electronic device, or stored in the memory of an electronic device in software form, so that the processor can call and execute the corresponding operations of each module.

[0099] In some exemplary embodiments of this application, an electronic device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, the electronic device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores a standard semantic library for the enterprise management building's digital twin. The network interface is used for communication with external devices via a network connection. When the computer program is executed by the processor, it implements a cross-domain heterogeneous data processing method for the enterprise management building's digital twin.

[0100] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0101] In some exemplary embodiments of this application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of a cross-domain heterogeneous data processing method for a digital twin of an enterprise management building as described in any of the exemplary embodiments above.

[0102] In some exemplary embodiments of this application, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of a cross-domain heterogeneous data processing method for a digital twin of an enterprise management building as described in any of the exemplary embodiments above.

[0103] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0104] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0105] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.< / extensionelements> < / bpmn:endevent> < / bpmn:startevent> < / bpmn:parallelgateway> < / bpmn:exclusivegateway> < / bpmn:task> < / bpmn:endevent> < / bpmn:startevent> < / bpmn:exclusivegateway> < / bpmn:task> < / extensionelements>

Claims

1. A method for cross-domain heterogeneous data processing of enterprise management building digital twins, characterized in that, The method includes: Obtain the standard semantic library of the enterprise management building digital twin. The standard semantic library configures the standard definitions and standard expressions of multiple management atoms of the full-domain process of the enterprise management building. The management atom is used to represent the node attributes of each process node of the business process in the full-domain process. The standard definition is used to describe the semantic connotation of the management atom. The standard expression is used to specify the unified expression form of the management atom. Receive cross-domain heterogeneous process data from different management systems; Obtain system prompt words for the large language model. The system prompt words are used to instruct the large language model to output a business process model file containing multiple process nodes, and record the node attributes of each process node in the business process model file. Input the system prompts and the cross-domain heterogeneous process data into the large language model to obtain the business process model file output by the large language model; Parse the business process model file and extract the node attributes of each process node from the business process model file; The node attributes of each process node are mapped to the standard definitions and standard expressions corresponding to each management atom in the standard semantic library, so as to complete the semantic fusion of cross-domain heterogeneous process data of different management systems.

2. The method according to claim 1, characterized in that, The step of extracting node attributes for each process node from the business process model file includes: Traverse the task nodes, gateway nodes, and event nodes in the business process model file; Extract the node attributes of each process node in the task node, the gateway node, and the event node respectively. The node attributes include step name, business object, business indicator, business rule, business risk, and business role. The process of mapping the node attributes of each process node to the standard definitions and standard expressions corresponding to each management atom in the standard semantic library includes: Map the step name of the task node to the standard definition of the management atom corresponding to the task node in the standard semantic library, and map the business object, business indicator, business rule, business risk and business role of the task node to the standard expression of the management atom corresponding to the task node; Map the step names of the gateway nodes to the standard definitions of the management atoms corresponding to the gateway nodes in the standard semantic library, and map the business rules of the gateway nodes to the standard expressions of the management atoms corresponding to the gateway nodes; The step name of the event node is mapped to the standard definition of the management atom corresponding to the event node in the standard semantic library, and the business rule of the event node is mapped to the standard expression of the management atom corresponding to the event node.

3. The method according to claim 2, characterized in that, The system prompt is used to instruct the large language model to output a business process model file containing multiple process nodes in an XML structure, in accordance with the BPMN specification. The traversal of task nodes, gateway nodes, and event nodes in the business process model file includes: Based on the XML structure and predefined node tags of the business process model file, the task nodes, gateway nodes, and event nodes in the business process model file are traversed.

4. The method according to claim 2, characterized in that, The step of extracting the node attributes of each process node in the task node, the gateway node, and the event node includes: Extract the step names of each process node from the inherent basic attributes of the task node, the gateway node, and the event node respectively; From the extended elements of the task node, extract the predefined business object attribute field, business indicator attribute field, business rule attribute field, business risk attribute field, and business role attribute field respectively to obtain the business object, business indicator, business rule, business risk, and business role of the task node; When traversing to the gateway node, the element content of the expression element of the gateway judgment condition is read from the sequence stream of the gateway node, and the read element content is used as the business rule of the gateway node. When traversing to the event node, the trigger type and trigger content are parsed from the event definition sub-element of the event node, and the business rules of the event node are constructed based on the trigger type and the trigger content.

5. The method according to claim 4, characterized in that, When traversing to the event node, the trigger type and trigger content are parsed from the event definition sub-element of the event node, and the business rules of the event node are constructed based on the trigger type and the trigger content, including: When traversing to the start event of the event node, the child elements of the start event are parsed. If the trigger type of the start event is identified as a message trigger type based on the child elements of the start event, the message content is extracted from the trigger content of the child elements of the start event, and the business rules of the start event are constructed based on the message trigger type and the message content. If the trigger type of the start event is identified as a timed trigger type based on the child elements of the start event, the time period is extracted from the trigger content of the child elements of the start event, and the business rules of the start event are constructed based on the timed trigger type and the time period. If the trigger type of the start event is identified as a conditional trigger type based on the child elements of the start event, the conditional expression is extracted from the trigger content of the child elements of the start event, and the business rules of the start event are constructed based on the conditional trigger type and the conditional expression. When traversing to the end event of the event node, the child elements of the end event are parsed. If the trigger type of the end event is identified as a termination type based on the child elements of the end event, the process state is extracted from the trigger content of the child elements of the end event. The business rules of the end event are constructed based on the termination type and the process state. If the trigger type of the end event is identified as a message output type based on the child elements of the end event, the output message is extracted from the trigger content of the child elements of the end event. The business rules of the end event are constructed based on the message output type and the output message.

6. The method according to claim 1, characterized in that, The process of mapping the node attributes of each process node to the standard definitions and standard expressions corresponding to each management atom in the standard semantic library includes: The semantic vectors of the node attributes of each process node are processed by semantic vectorization through a pre-trained domain word embedding model. The semantic vectors of the node attributes of each process node are compared with the semantic vectors in the semantic vector library of the standard semantic library for similarity calculation. Each semantic vector in the semantic vector library corresponds to a management atom in the standard semantic library. Management atoms with similarity greater than the similarity threshold are used as the management atoms corresponding to the node attributes of each process node; Map the node attributes of each process node to the corresponding management atomic standard definition and standard expression.

7. The method according to claim 6, characterized in that, The method further includes: Obtain a historical mapping dataset, which contains multiple pairs of matching node attributes and management atoms calculated through similarity; Based on the historical mapping dataset, the distributional characteristic parameters of similarity are calculated through statistical analysis, and the distributional characteristic parameters include the mean and standard deviation; The similarity threshold is determined based on the mean and standard deviation.

8. A cross-domain heterogeneous data processing system for enterprise management building digital twins, characterized in that, The system includes: The standard semantic library acquisition module is used to acquire the standard semantic library of the enterprise management building digital twin. The standard semantic library configures the standard definitions and standard expressions of multiple management atoms of the full-domain process of the enterprise management building. The management atom is used to represent the node attributes of each process node in the business process in the full-domain process. The standard definition is used to describe the semantic connotation of the management atom, and the standard expression is used to specify the unified expression form of the management atom. The receiving module is used to receive cross-domain heterogeneous process data from different management systems; The system prompt word acquisition module is used to acquire system prompt words of the large language model. The system prompt words are used to instruct the large language model to output a business process model file containing multiple process nodes, and to record the node attributes of each process node in the business process model file. The business process model file acquisition module is used to input the system prompt words and the cross-domain heterogeneous process data into the large language model to obtain the business process model file output by the large language model. The node attribute extraction module is used to parse the business process model file and extract the node attributes of each process node from the business process model file. The data mapping module is used to map the node attributes of each process node to the standard definitions and standard expressions corresponding to each management atom in the standard semantic library, so as to complete the semantic fusion of cross-domain heterogeneous process data of different management systems.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images