A method for integrating digital delivery information for cleanroom renovation projects

By generating component-level semantic state snapshot chains and performing difference-aware processing in cleanroom renovation projects, the configuration maintenance and format adaptation issues in multi-source data integration are solved, improving the real-time performance and accuracy of data access and enhancing the responsiveness to changing scenarios.

CN123115251APending Publication Date: 2026-11-17GUANGDONG HENGTAI DESIGN ENG CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611245713.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-17
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies for integrating multi-source heterogeneous data in the digital delivery of cleanroom renovation projects suffer from cumbersome configuration and maintenance, as well as inefficient format adaptation. In particular, they are slow to respond to sudden and high-frequency changes, making it difficult to achieve real-time mapping and consistency assurance.

Method used

By acquiring key component structural definition information, business constraint rules, and delivery requirement data from the BIM model, a component-level semantic state snapshot is generated, and a snapshot chain is constructed. The latest semantic state snapshot is located using a composite index key, difference fields are identified, a semantic mapping instruction set is generated, and atomic transformation processing is performed to achieve data integration.

Benefits of technology

It significantly improves the real-time performance and accuracy of data access, reduces human configuration errors, enhances the responsiveness to upgrades and equipment replacements, and ensures semantic robustness and consistency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123115251A_ABST
    Figure CN123115251A_ABST
Patent Text Reader

Abstract

This invention relates to a method for integrating digital delivery information in cleanroom renovation projects. Addressing the challenges of complex data sources, heterogeneous structures, inconsistent semantics, and difficulties in data traceability, this method automatically extracts structural and constraint information of key components, achieving semantic alignment and standardization across multiple data sources. It dynamically responds to semantic differences and automatically generates correction instructions, ensuring data consistency and meeting business integration needs in edge environments. Ultimately, it constructs an immutable data traceability certificate, improving the traceability and consistency of project data. This provides effective technical support for data management and acceptance processes in cleanroom renovation and complex BIM scenarios, offering significant advantages in real-time performance, scalability, and security.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of cleanroom engineering information integration and BIM multi-source data fusion technology, and in particular to a method for integrating digital delivery information for cleanroom renovation projects. Background Technology

[0002] Existing multi-source heterogeneous data integration technologies in the digital delivery of cleanroom renovation projects typically use Building Information Modeling (BIM) as the core data hub, driving cross-system data fusion by constructing a complete engineering semantic model. Current mainstream information integration modules focus on the following solutions: First, based on a unified data platform, multi-source data acquisition and centralized ETL (Extract-Transform-Load) processes convert all external data from sensors, construction progress, design drawings, etc., into preset standard formats before mapping them into the BIM database; second, using rule engines, mapping tables, and other methods to achieve the verification and adaptation of various structures, field attributes, units, and other common specifications. These existing solutions are suitable for digital delivery scenarios such as large and medium-sized cleanrooms, hospital purification areas, and microelectronic clean spaces, enabling automatic conversion and data collection of basic information flows in the early stages of a project.

[0003] With the rapid evolution of digitalization across industries, the rapid progress of project delivery, and the widespread adoption of heterogeneous intelligent terminals, data types in the cleanroom construction engineering field are becoming increasingly diverse. For example, on-site sensors frequently go online and offline; data structures and descriptive granularities differ significantly among different construction contractors; on-site logs contain multilingual and semi-structured text; and even updated drawings often have multiple versions coexisting. Existing multi-source data fusion technologies are mostly based on a "full standardization, post-batch processing" paradigm, requiring mandatory format conversion of all heterogeneous data sources. This not only leads to cumbersome configuration and maintenance and inefficient format adaptation but also causes the following problems: large-scale field mapping, structural decomposition, and semantic completion operations are required when data flows in; if new types of data sources are encountered, it is often necessary to write dedicated protocol parsers or extend the rule engine, severely restricting the system's response speed to sudden, high-frequency changes. For example, after the cleanroom air filtration system is upgraded, new monitoring parameters are generated, and existing systems often struggle to achieve real-time mapping and consistency assurance. Summary of the Invention

[0004] This application provides a method for integrating digital delivery information for cleanroom renovation projects, aiming to solve the problems of cumbersome configuration and maintenance and inefficient format adaptation in the digital information integration of cleanroom renovation projects mentioned in the background art.

[0005] This application provides a method for integrating digital delivery information for cleanroom renovation projects, specifically including: S1: Obtain the structural definition information, business constraint rules, and delivery requirement data of key components in the BIM model of the cleanroom decoration project; generate component-level semantic state snapshots based on the structural definition information, business constraint rules, and delivery requirement data; and combine multiple component-level semantic state snapshots arranged in time series to construct a component-level snapshot chain. S2: Receive multi-source data streams from sensors, construction logs, or design drawings; construct a composite index key based on the multi-source data streams; and use the composite index key to locate and obtain the latest component-level semantic state snapshot in the component-level snapshot chain. S3: Compare the multi-source data stream with the latest component-level semantic state snapshot, identify the set of discrepancies in field name, unit, value range or time precision, and generate a difference perception result based on the set of discrepancies. S4: Process the difference perception result to obtain a set of field path redirection instructions, a unit conversion coefficient field, and an enumeration value mapping table, and encapsulate the set of field path redirection instructions, the unit conversion coefficient field, and the enumeration value mapping table into a semantic mapping instruction set; S5: Load the semantic mapping instruction set to the edge computing node for execution, perform atomic transformation processing on the multi-source data stream, generate standardized fused data, and complete the integration of digital delivery information for engineering projects.

[0006] The digital delivery information integration method for cleanroom renovation projects provided in this application has the following beneficial effects: This solution binds semantic snapshots containing structural definitions and business constraints to key components during the modeling phase, forming a state snapshot chain. This allows data access to be performed by simply locating the latest snapshot and activating a difference-aware engine to automatically identify deviations in field names, units, and enumeration values. This precisely converges the transformation actions to a minimum impact range, significantly improving the real-time performance and accuracy of data access. Simultaneously, it clearly defines path redirection, unit conversion, and enumeration mapping rules, embedding them directly into edge devices for local execution. This eliminates the need for remote scheduling or manually written parsing code. All transformation results are accompanied by an immutable traceability credential containing the component ID, snapshot version hash, and digital signature, and are synchronously written to the component-level log, completely eliminating the risk of human configuration errors. New data sources can be automatically adapted across the entire chain simply by updating the semantic snapshot, greatly improving the responsiveness and semantic robustness to on-site changes such as upgrades and equipment replacements. Attached Figure Description

[0007] Figure 1 This is the main flowchart of a digital delivery information integration method for cleanroom renovation projects.

[0008] Figure 2 This is a sub-flowchart of a digital delivery information integration method for cleanroom renovation projects.

[0009] Figure 3 This is another sub-flowchart of a digital delivery information integration method for cleanroom renovation projects. Detailed Implementation

[0010] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0011] The following disclosure provides many different embodiments or examples for implementing different structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0012] like Figure 1 As shown, this application provides a method for integrating digital delivery information for cleanroom renovation projects, specifically including: S1: Obtain the structural definition information, business constraint rules, and delivery requirement data of key components in the BIM model of the cleanroom decoration project. Generate component-level semantic state snapshots based on the structural definition information, business constraint rules, and delivery requirement data, and combine multiple component-level semantic state snapshots arranged in time sequence to construct a component-level snapshot chain.

[0013] S2: Receive multi-source data streams from sensors, construction logs, or design drawings, construct a composite index key based on the multi-source data streams, and use the composite index key to locate and obtain the latest component-level semantic state snapshot in the component-level snapshot chain.

[0014] S3: Compare the multi-source data stream with the latest component-level semantic state snapshot to identify the set of discrepancies in field name, unit, value range or time precision, and generate a difference perception result based on the set of discrepancies.

[0015] S4: Process the difference perception result to obtain a set of field path redirection instructions, a unit conversion coefficient field, and an enumeration value mapping table, and encapsulate the set of field path redirection instructions, the unit conversion coefficient field, and the enumeration value mapping table into a semantic mapping instruction set.

[0016] S5: Load the semantic mapping instruction set to the edge computing node for execution, perform atomic transformation processing on the multi-source data stream, generate standardized fused data, and complete the integration of digital delivery information for engineering projects.

[0017] Step S1: Obtain structural definition information, business constraint rules, and delivery requirement data of key components in the BIM model of the cleanroom renovation project; generate component-level semantic state snapshots based on the structural definition information, business constraint rules, and delivery requirement data; and combine multiple component-level semantic state snapshots arranged in time series to construct a component-level snapshot chain. Specifically, this includes: S1.1: Based on the geometric topology in the cleanroom decoration project BIM model, semantic entity recognition processing is performed on key components such as high-efficiency filters, airtight doors, and differential pressure sensor installation points to obtain structural definition information containing industrial basic class type identifiers and attribute set IDs.

[0018] Load the BIM model file of the cleanroom renovation project, parse the industrial basic category standard data flow, and extract the geometric topology and spatial hierarchy. Traverse the model entity tree and, based on the pre-set cleanroom key component classification rule library, filter target objects such as HEPA filters (FFUs), airtight doors, and differential pressure sensor installation points. Perform semantic feature extraction on the selected component instances, read their industrial basic category Type attribute values, and determine their type identifier in the industrial basic category standard. Associate the component's attribute set definition and obtain Pset_ID to lock specific business attribute groups. Perform digest calculation on the combination of the component's geometric GUID and attribute set ID using a hash method to generate a unique structural fingerprint. Map the structural fingerprint to an internal index table to establish a fast retrieval path from physical entities to logical semantic nodes. Verify the completeness of the extracted structural definition information to ensure that it includes necessary spatial attribution and connection relationship data.

[0019] Through the above processing method, the raw BIM data from the previous step is transformed into structural definition information containing industrial basic type identifiers and attribute set IDs, enabling accurate identification and standardized description of semantic entities of key components, and providing definite structural anchor points for subsequent business rule binding.

[0020] S1.2: Using the attribute set ID in the structure definition information, associate the industry standard library with the project delivery standard, and perform business rule instantiation processing on each key component to generate business constraint rules that include unit constraints, numerical threshold ranges and enumerated status lists.

[0021] The system receives the structural definition information, including the industrial basic class type identifier and attribute set ID, output from S1.1, as input for business rule instantiation. It parses the classification coding field in the attribute set ID, maps it to a pre-set industry standard knowledge base index table, and retrieves national standards for cleanroom construction and project-specific delivery standard documents matching the component type. It extracts definition clauses regarding physical dimensions, numerical thresholds, and state enumerations from the standard documents, constructing an original constraint rule template for the current component type. It performs parameterized instantiation on the original constraint rule template, replacing variables in the general standards with specific design indicators for the current engineering project, generating business constraint rules containing unit constraints, numerical threshold ranges, and enumerated state lists. By linking the industry standard library with project delivery standards, the structural definition information from the previous step is transformed into business constraint rules with legal force and engineering applicability, achieving a deep extension of component semantics from geometric form to business logic.

[0022] For example, for the differential pressure sensor mounting point component, its attribute set ID is mapped to the GB 50457-2019 standard. The system extracts the "differential pressure between adjacent clean areas" clause, instantiates it as a unit constraint "Pa", a numerical threshold range "≥15", ​​and an enumerated state list "normal / alarm". If the design target is set to 20Pa, then specific business constraint rules are generated: the unit must be Pa, the normal range is [15, 25], and the state enumeration value is normal or alarm. This process ensures that subsequent data comparisons have a clear quantitative benchmark, significantly improving the logical rigor of data fusion.

[0023] S1.3: Based on the mapping relationship between the business constraint rules and the engineering acceptance process, the document reference path and deliverable association logic of key components are encapsulated to form delivery requirement data containing read-only attribute binding requirements.

[0024] Based on the business constraint rules generated by S1.2, the engineering acceptance process knowledge base is invoked to retrieve a list of deliverable templates matching the industrial foundation class type identifier of the current key component. This list includes templates for test reports, material certificates, and commissioning records required under the ISO 14644 standard for components such as cleanroom HEPA filters and airtight doors. The retrieved deliverable templates are structured and parsed, extracting the metadata field definitions for each template, including document type codes, required attribute fields, associated process WBS nodes, and version control requirements. These metadata fields are mapped to attribute constraints of read-only objects in the industrial foundation class of the BIM model. A document reference path generation method is constructed, automatically generating a standardized file storage directory tree structure based on the spatial coordinates of the component and its system affiliation. This structure ensures that multiple documents generated for the same component at different construction stages can be archived in chronological order, avoiding path conflicts. The generated directory tree structure is bound to the metadata fields of the deliverable templates to form delivery requirement data containing read-only attribute binding requirements. This object clarifies the path format and association logic that subsequent accessed construction logs or acceptance reports must meet. By using the above processing method, the business rules of the previous step are transformed into executable document association contracts, realizing the automated and accurate mounting of multi-source unstructured documents and BIM components, and significantly improving the retrieval efficiency and completeness of delivery data.

[0025] For example, for a HEPA filter (FFU) component in a certain area of ​​a cleanroom, the system retrieves the HEPA leak test report and airflow balance test record that need to be associated. The method generates a storage path " / ProjectA / Zone3 / FFU-001 / Commissioning / " and maps the "Initial Pressure Difference" field in the report to the BIM attribute Pset_FilterCommon.InitialPressureDrop. When optical character recognition identifies a file named "FFU-001_LeakTest.pdf", the system automatically links it to the read-only attribute of that component, without manual intervention, ensuring a one-to-one correspondence between the delivered data and the model entity.

[0026] S1.4: Integrate the structural definition information, business constraint rules and delivery requirement data, and use a serialization method to generate a component-level semantic state snapshot containing a unique version hash value and a high-precision timestamp, so as to realize the conversion of the component's static geometric information into a dynamic semantic contract executor.

[0027] Receive structural definition information, business constraint rules, and delivery requirement data from the output of steps S1.1 to S1.3 as the original input data source for building component-level semantic state snapshots.

[0028] The JSON-LD serialization protocol is used to encapsulate the above three types of heterogeneous data in a structured manner. The industrial basic class type identifier and attribute set ID are mapped to standard schema nodes, the unit constraints, numerical thresholds and enumeration lists are serialized into a validation rule array, and the read-only binding logic is encapsulated into a reference pointer object, forming a unified intermediate semantic description package.

[0029] The serialized semantic description package is subjected to normalized sorting processing. The key-value pairs are rearranged according to the lexicographical order of the field names to eliminate the differences in byte streams caused by different data writing orders, and to ensure that the same semantic content has a completely consistent binary representation at different generation times.

[0030] Based on the normalized binary byte stream, the SHA-256 hash method is called to calculate the data fingerprint and generate a unique version hash value with a length of 64 characters. This hash value serves as a digital digest of the current snapshot content and is used for subsequent data integrity verification and version comparison.

[0031] Obtain the current timestamp output by the system's high-precision clock source, accurate to the nanosecond level, and combine it with the generated unique version hash value and the component's globally unique identifier GUID to construct metadata header information containing spatiotemporal dimensions.

[0032] The metadata header information is then concatenated with the normalized semantic description package to generate an immutable component-level semantic state snapshot file. This snapshot fully records the static geometric attributes, dynamic business constraints, and delivery-related logic of the component at a specific moment.

[0033] Through the serialization, normalization, hash calculation and encapsulation processes described above, the scattered structure, constraint and delivery data from the previous step are transformed into component-level semantic state snapshots with unique identification and tamper-proof characteristics. This realizes the essential transformation of components from static geometric models to dynamic semantic contract executors, providing standardized data units for the subsequent construction of snapshot chains that support historical backtracking.

[0034] For example, for a cleanroom high-efficiency particulate air (HEPA) filter (FFU) component, its industrial base class type is integrated as FlowTerminal, the attribute set ID is Pset_FilterCommon, the business constraint specifies a differential pressure threshold of 15Pa to 50Pa, and the delivery requirement is associated with the acceptance report Doc_001. The system first serializes the above information into JSON-LD format, and generates a binary stream after lexicographically sorting the key names. The SHA-256 method is called to calculate the hash value a1b2...c3d4. Combined with the timestamp 17156688000000000000 and the component GUID, the final snapshot object is generated. The size of this snapshot file is controlled within 2KB, and the generation time is less than 5ms, ensuring the efficiency and consistency of generating snapshots of tens of thousands of components in a large-scale BIM model, providing an accurate benchmark anchor point for subsequent incremental semantic mapping.

[0035] S1.5: Based on the timestamp sequence of the component-level semantic state snapshots, perform chain linking processing on multiple component-level semantic state snapshots of the same key component at different engineering stages to construct a component-level snapshot chain that supports historical backtracking and difference comparison.

[0036] Obtain the component-level semantic state snapshot set generated by S1.4, which includes a unique version hash and a high-precision timestamp, as the basic data unit for constructing the snapshot chain. For multiple semantic state snapshots of the same key component generated at different engineering stages, extract the high-precision timestamp field they carry, and use a quicksort method to sort the snapshot set in ascending order according to time sequence, establishing a linear logical relationship between snapshots in the time dimension. Traverse the sorted snapshot sequence, calculate the semantic difference between two adjacent snapshots, and identify state nodes that have undergone substantial changes by comparing the hash value changes of structural definition information, business constraint rules, and delivery requirement data. Based on the version hash dependency relationship between adjacent snapshots, establish a forward pointer linking mechanism, using the version hash value of the current snapshot as the predecessor index identifier of the next snapshot, forming a unidirectional and irreversible chain data structure. Write the component's globally unique identifier and initial creation timestamp into the chain head node, and mark the latest valid state identifier at the chain tail node to ensure the integrity and traceability of the chain list structure. A Merkle Tree hash verification mechanism is introduced, performing SHA-256 hash operations on the complete payload data of each node in the linked list to generate a node fingerprint. This fingerprint is then incorporated into the metadata header of the next node, achieving tamper-proof data lineage binding. Through the above chain-linking process, discrete component state snapshots are transformed into a component-level snapshot chain that supports historical backtracking, version comparison, and integrity verification, enabling dynamic updates of semantic benchmarks and audit trail capabilities during multi-source data fusion.

[0037] For example, for the FFU component of the high-efficiency filter in a clean room, the system generates an initial snapshot Hash_A at time T1, and generates a snapshot Hash_B at time T2 after adjusting the wind speed threshold due to design changes. The system extracts timestamps to determine that T1<T2, calculates the difference between Hash_A and Hash_B, and confirms that the business constraint rule has changed. A pointer from Hash_A to Hash_B is established, and the SHA-256 value of Hash_A is embedded into the metadata header of Hash_B. When the sensor is connected at time T3, the system quickly locates Hash_B along the snapshot chain as the latest semantic benchmark, ensuring that data mapping is executed based on the latest design specifications, which significantly improves the efficiency of data consistency verification.

[0038] As shown Figure 2 , the step S2: receiving multi-source data streams from sensors, construction logs or design drawings, constructing a composite index key according to the multi-source data streams, and using the composite index key to locate and obtain the latest component-level semantic state snapshot in the component-level snapshot chain.

[0039] This step is responsible for receiving multi-source data streams from sensors, construction logs or design drawings, extracting the engineering context identifier carried therein, using the identifier to perform precise positioning and matching in the pre-built component-level snapshot chain, so as to obtain the latest component-level semantic state snapshot corresponding to the current data stream, and provide a unique semantic benchmark anchor point for subsequent differential comparison and incremental mapping.

[0040] S2.1: Perform protocol parsing and payload extraction processing on multi-source data streams such as sensor data packets, construction log text fragments and optical character recognition results of design drawings that enter the information integration module, so as to separate the payload content containing original business data and the metadata header information carrying source features.

[0041] Sub-step S2.1 aims to perform protocol parsing and payload extraction on the multi-source data streams entering the information integration module, separating the original business payload and source feature metadata. The system receives data streams from sensor MQTT messages, construction log TXT text, and design drawing optical character recognition (OCR) JSON results. For sensor data, the MQTT header is parsed, and the device ID and timestamp at the Topic level are extracted as metadata. The payload is retained as the raw reading payload in binary or JSON format. For construction log text, a natural language processing preprocessing module is used to identify unstructured paragraphs in the text, extracting the main text containing process descriptions and personnel signatures as the payload, while capturing file system attributes such as file creation time and author ID as metadata. For design drawing OCR results, the structured JSON output is parsed, and the identified text coordinates, confidence scores, and layer information are encapsulated into a metadata header. The specific labeled text content is extracted as the payload. Through a unified interface adapter, the metadata from the above different sources is standardized into a general metadata header object containing data source type identifiers, original encoding formats, and acquisition time windows. This processing method transforms the multi-source raw data from the previous step into a well-structured binary tuple of "metadata header + business payload," achieving initial decoupling and standardized entry point for heterogeneous data sources. This provides a clean and labeled data foundation for the accurate extraction of subsequent engineering context identifiers.

[0042] For example, the system receives an MQTT data packet from a cleanroom differential pressure sensor, with the Topic "cleanroom / zoneA / sensor001" and the Payload {"val": 15.2, "unit": "Pa"}. The parser extracts "sensor001" as the device ID, "1715623400" as the timestamp and stores it in the metadata header, and {"val": 15.2, "unit": "Pa"} as the payload. Simultaneously, a construction log "FFU installation complete.txt" is received, and the filename, modification time 2024-05-13 10:00:00, and author "Worker_A" are extracted as metadata. The body text "FFU-01 installation completed, sealing check passed" is the payload. The system outputs a standardized binary object, with the metadata header containing source_type="sensor", device_id="sensor001", timestamp=1715623400, and the payload being the original JSON string. This process ensures that subsequent steps can be context-based based on unified metadata fields, significantly improving the parsing efficiency and accuracy of multi-source data access.

[0043] S2.2: Based on the floor code field, process WBS code field and equipment serial number field in the metadata header information, a multi-dimensional key-value concatenation method is used to perform engineering context identifier generation processing to construct a composite index key that can uniquely point to a specific BIM component instance.

[0044] Receive the metadata header information parsed by S2.1, and extract the floor code field, process WBS code field and equipment serial number field as the original index elements.

[0045] The extracted floor codes are padded with fixed length to convert them into a fixed-length string format, eliminating sorting ambiguity caused by differences in the number of bits.

[0046] The process WBS code is expanded hierarchically, and based on the project decomposition structure standard, the multi-level WBS code is mapped to a unique one-dimensional linear identifier.

[0047] A hash digest calculation is performed on the device serial number, using the SHA-256 method to generate a fixed-length hash value, ensuring the uniqueness of the serial number and protecting privacy. A multi-dimensional key-value concatenation method is constructed, and a formula for generating composite index keys is defined:

[0048] in, For composite index keys, For integration, For standardized floor coding, For one-dimensional process identification, This is a hash digest of the device serial number.

[0049] Perform a string concatenation operation, connecting the three elements—floor, process, and equipment—into a single string according to their priority order.

[0050] The generated composite index key undergoes an integrity check to verify its length conforms to preset specifications, ensuring its uniqueness and retrieval capability. Through multi-dimensional key-value concatenation, scattered multi-source metadata is transformed into a composite index key uniquely pointing to a specific BIM component instance, achieving precise anchoring of heterogeneous data to BIM model components and providing efficient retrieval basis for rapid location of subsequent snapshot chains.

[0051] For example, a differential pressure sensor in a cleanroom reports data with the floor code "F3", the process WBS code "HVAC-02-05", and the equipment serial number "SN-998877". The floor code "F3" is padded to "F03". The WBS code "HVAC-02-05" is mapped to the linear identifier "H0205". The SHA-256 hash value of the serial number "SN-998877" is calculated, and the first 8 bits are taken as "a1b2c3d4". These are then concatenated to generate a composite index key "F03_H0205_a1b2c3d4". This index key uniquely points to a specific differential pressure sensor component in the HVAC 02-05 area on the third floor of the BIM model. The system completes the location within milliseconds, significantly improving data fusion efficiency.

[0052] S2.3: Using the composite index key as a query condition, perform hash addressing and version retrieval processing in the index database that stores the full set of component-level snapshots to lock the set of component-level semantic state snapshots of all historical versions of the target key component in the current time window.

[0053] Receive the composite index key generated by S2.2, which contains a hash digest of the floor code, process WBS code and equipment serial number, as a unique addressing identifier for database queries.

[0054] The hash ring routing method of the distributed key-value storage engine is invoked to map the composite index key to a specific data shard node, eliminating I / O latency caused by global lock contention.

[0055] Within the target shard, the B+ tree index structure is used to accurately match the component ID primary key, quickly locating the data page set containing all historical version snapshots of the component.

[0056] Traverse the version metadata area in the data page, extract the timestamp field and version hash value of each snapshot record, and build a temporary version candidate list.

[0057] Apply time window filtering logic to remove expired snapshots that are beyond the validity period of the current project stage, and retain only semantic state records that are within the active lifecycle.

[0058] Perform integrity checks on the remaining candidate snapshots to verify the consistency between their digital signatures and storage hashes, ensuring the immutability of the read data.

[0059] The verified snapshot pointers and their metadata are encapsulated into an ordered array to form a full historical version snapshot set of the target key component in the current time window.

[0060] Through the hash addressing and version retrieval processes described above, the composite index key from the previous step is transformed into a structured set of historical version snapshots, enabling millisecond-level precision in locking the semantic state of components and providing a complete data foundation for subsequent latest state determination.

[0061] For example, for a cleanroom HEPA filter (FFU) component, its composite index key is "L3-HVAC-FFU01". The system hashes and maps this key to the Shard-05 node. In Shard-05, a B+ tree search finds the data page corresponding to the component ID "Industrial Basic Class_FFU_8821". This page contains five historical version snapshots, with timestamps from T1 to T5. The system sets the current engineering phase as "Installation and Commissioning Period", with a valid time window of [T3, T5]. Snapshots from the early design phases T1 and T2 are removed from the filter. SHA-256 verification is performed on the three snapshots T3, T4, and T5 to confirm data integrity. The final output contains a set of three valid snapshots, each with a size controlled within 2KB, and the search time is less than 5ms, significantly improving the efficiency of context acquisition when accessing multi-source data.

[0062] S2.4: Based on the high-precision timestamp and unique version hash value carried by each snapshot in the component-level semantic state snapshot set, perform time-series sorting and latest state determination logic operations to filter out the latest component-level semantic state snapshot representing the latest semantic constraints from multiple historical versions.

[0063] The system receives a set of historical version semantic state snapshots of the target key component retrieved in step S2.3. This set contains multiple snapshot objects generated by the same component at different engineering stages, each carrying a high-precision timestamp and a unique version hash value. The system iterates through the snapshot set, extracting the timestamp field data from each snapshot object and converting the timestamps to a unified standard UTC time format to eliminate time inconsistencies caused by sensor clock drift or system time zone differences. Based on the converted standard UTC timestamps, the snapshot set is sorted in ascending order using a quicksort method, constructing a strictly increasing snapshot sequence in chronological order to ensure that subsequent version determination logic is based on the correct temporal linear relationship. In the sorted snapshot sequence, the snapshot object at the end is extracted as a candidate latest state, and its unique version hash value is read. This hash value, generated by the SHA-256 method based on the snapshot content, identifies the uniqueness of the version's semantic contract. A version hash integrity check is performed, comparing the candidate latest state hash value with the latest valid hash pointer recorded in the index database to confirm that the snapshot has not been tampered with and is the latest valid version recognized by the current system, eliminating interference from dirty data caused by network latency. If the hash verification passes, the candidate snapshot is locked as the final latest component-level semantic state snapshot; if the verification fails, the process is repeated back to the next newest snapshot until the first snapshot object with a valid hash match is found. Through a dual verification mechanism of temporal sorting and hashing, the component-level semantic state snapshot representing the latest semantic constraints is accurately selected from multi-version historical data, ensuring the real-time performance and consistency of the data fusion benchmark.

[0064] For example, for a cleanroom HEPA filter (FFU) component, the system retrieves five historical snapshots, timestamps T1 to T5. After UTC normalization, T5 is the most recent time point. The SHA-256 hash value H5 of the snapshot corresponding to T5 is extracted and compared with the database pointer. The system directly loads the T5 snapshot as the baseline, which includes a wind speed threshold constraint of 0.45 m / s. If H5 does not match the pointer, the system automatically backtracks to verify the T4 snapshot, ensuring the absolute reliability of the fused baseline, with processing time controlled within 2 ms.

[0065] S2.5: Load the latest component-level semantic state snapshot into the context register of the memory difference awareness engine and establish a temporary mapping association between it and the original multi-source data stream to complete the preparation of benchmark data for subsequent item-by-item semantic comparison.

[0066] The system receives the latest component-level semantic state snapshot binary data stream filtered by S2.4 and loads it into the dedicated context register area of ​​the difference-aware engine using memory-mapped file technology, achieving zero-copy data access. The loaded snapshot data is deserialized and parsed to extract the industrial basic class type identifier, attribute set ID, and business constraint rules, constructing a standardized semantic contract object model in memory. A temporary pointer association is established between this semantic contract object and the original multi-source data stream, forming a logical comparison benchmark domain. A multi-dimensional semantic verification buffer is initialized, pre-allocating memory space for storing field name hash values, unit dimension codes, and enumeration mapping tables. Through memory address alignment optimization technology, key constraint parameters in the snapshot are cached in the CPU L1 cache line, reducing memory latency during subsequent item-by-item comparisons. A Bloom filter is used to quickly determine whether fields in the original data stream exist in the allowed field set defined in the snapshot, eliminating invalid and redundant fields. The numerical threshold ranges in the snapshot are converted into an interval tree structure to accelerate the efficiency of subsequent numerical boundary detection. Through the above processing method, the latest snapshot obtained in the previous step is transformed into a semantic comparison benchmark environment that can be accessed in memory in real time, realizing microsecond-level data preparation response and providing a low-latency execution basis for the high-concurrency semantic consistency verification of step S3.

[0067] For example, for a cleanroom HEPA filter (FFU) component, its latest semantic state snapshot size is 256 bytes. The system maps this snapshot to a memory address starting at 0x7F00 using mmap. The parsed industrial base class type is FlowTerminal, and the attribute set ID is Pset_FFUCleanroom. A range tree containing differential pressure thresholds of 15Pa-50Pa is constructed. The raw sensor data stream JSON packets are pointers to different offsets within the same memory page. The Bloom filter confirmation field pressure_kpa exists in the allowed set. Range tree construction takes less than 1 microsecond. The L1 cache hit rate reaches over 98%. Finally, the baseline environment is set up within 5 microseconds. Compared to traditional database query methods, data preparation latency is reduced by two orders of magnitude, significantly improving real-time fusion efficiency.

[0068] Step S3: Compare the multi-source data stream with the latest component-level semantic state snapshot to identify the set of discrepancies in field name, unit, value range or time precision, and generate a difference perception result based on the set of discrepancies.

[0069] This step is responsible for performing a deep comparison between the metadata of multi-source data streams and the expected semantics in the component-level semantic state snapshot. By building a difference-aware engine, it identifies deviations in dimensions such as field name, unit, value range, and time precision, and generates a set of difference fields and difference-aware results. This provides accurate decision-making basis for the automatic generation of subsequent micro-execution instructions and is the core logical link to achieve zero-redundancy incremental fusion.

[0070] S3.1: Based on the engineering context identifier extracted from the multi-source data stream, retrieve the latest component-level semantic state snapshot from the component-level snapshot chain, and perform deserialization parsing on the structural definition information, business constraint rules and delivery requirement data contained in the latest component-level semantic state snapshot to obtain a standardized semantic contract set containing industrial basic class type identifier, attribute set ID, unit constraint, numerical threshold range, enumerated state list and read-only attribute binding requirements.

[0071] The system receives the latest component-level semantic state snapshot binary data stream from step S2. This data stream contains serialized structural definitions, business constraints, and delivery requirements. A lightweight deserialization engine is invoked to parse the binary data stream according to the predefined Protocol Buffers schema, restoring it to an in-memory object graph structure. Industrial base class type identifiers and attribute set IDs are extracted from the object graph to construct a component geometric semantic index and establish the target namespace for data mapping. Business constraint rule blocks are read, and the encapsulated unit system definitions, valid numerical range boundaries, and enumerated state dictionaries are parsed to form a dimensional verification benchmark table. Delivery requirement data is loaded, and read-only attribute binding paths and associated document type restrictions are extracted to generate a metadata mounting rule set. The geometric semantic index, dimensional verification benchmark table, and metadata mounting rule set obtained from the above parsing are combined and assembled into a standardized semantic contract set object. Through deserialization parsing, the binary snapshot from the previous step is transformed into a structured contract set containing complete semantic constraints, achieving semantic activation from static storage state to dynamic execution state, providing an accurate comparison benchmark for subsequent multi-dimensional consistency verification.

[0072] For example, the latest semantic state snapshot of a cleanroom HEPA filter (FFU) component has a binary payload length of 256 bytes. The deserialization engine identifies the header identifier 0x8950, extracts the industrial basic class type as FlowTerminal based on the schema definition, and assigns the attribute set ID Pset_FilterCommon. The business constraint block is parsed to obtain the pressure unit constraint as Pa, the numerical threshold range as [15, 50], and the enumerated state list as {running: 1, stopped: 0, alarm: 2}. Delivery requirements are extracted, determining that the acceptance report must be bound to a read-only field, and the format is limited to PDF. The final generated semantic contract set contains 12 specific constraint indicators across 3 categories, with a memory footprint of only 4KB and a parsing time of less than 2ms, ensuring low-latency startup of subsequent comparison processes.

[0073] S3.2: Using the standardized semantic contract set as a mapping benchmark, metadata extraction and normalization preprocessing are performed on the original data packets in the multi-source data stream to separate the field name string to be compared, the original unit of measurement identifier, the real-time collected value range and the timestamp accuracy level, thereby forming a structured set of meta-information to be compared.

[0074] The system receives raw data packets from multi-source data streams and calls the corresponding parsing adapter based on the data source protocol type. For JSON-formatted sensor data, it performs key-value pair traversal operations to extract field name strings and their associated raw values. For unstructured construction log text, it uses a natural language processing model to identify key entities and separate status descriptors and timestamps. For optical character recognition results from design drawings, it extracts dimensional parameters and unit identifiers using regular expression matching.

[0075] Perform dimensional standardization mapping on the extracted original unit identifiers. Establish a unit conversion dictionary to uniformly convert non-standard units such as kPa and mmHg to the Pa standard unit. Calculate unit conversion coefficients to ensure that subsequent numerical comparisons are based on the same dimensional system. For dimensionless enumerated state values, retain their original string form for subsequent semantic matching.

[0076] Analyze the timestamp precision level of real-time acquired data. Parse the ISO 8601 format time string to determine the time precision to the second, millisecond, or microsecond level. Compare with the minimum time granularity requirements defined in the component snapshot and mark data items with insufficient time precision. Extract the dynamic range of numerical changes and record the maximum and minimum values ​​to assess the range of data fluctuation.

[0077] Integrating the above processing results, a structured set of metadata to be compared is constructed. This set includes standardized field names, numerical values ​​with unified dimensions, normalized unit identifiers, and time precision labels. Each data item is appended with a source tracking ID to ensure data lineage traceability.

[0078] By extracting metadata and performing normalized preprocessing, multi-source heterogeneous raw data is transformed into a standardized set of metadata to be compared, eliminating the interference of format differences on semantic comparison and achieving a foundation for efficient and accurate difference identification.

[0079] For example, the system receives JSON data from a cleanroom differential pressure sensor: {"pressure_kpa": 15.2, "ts": "2023-10-01T10:00:00.123Z"}. It parses and extracts the field name "pressure_kpa", the value 15.2, and the unit "kPa". It then queries a unit dictionary to determine that 1 kPa = 1000 Pa, generating a conversion factor of 1000. The timestamp is parsed to ensure millisecond-level precision. If the component snapshot requires second-level precision and the unit is Pa, then the system is marked as having a unit deviation, but the time precision meets the requirements. The system outputs structured metadata: field name "pressure_kpa", original value 15.2, unit "kPa", conversion factor 1000, and time precision "ms". This process standardizes heterogeneous data, significantly improving subsequent comparison efficiency.

[0080] S3.3: Based on the structured set of meta-information to be compared and the standardized semantic contract set, a multi-dimensional semantic consistency verification method is initiated. This method performs the following steps: fuzzy matching degree calculation of field name strings, dimensional compatibility verification of original unit of measurement identifiers and unit constraints, boundary inclusion detection of real-time collected numerical ranges and numerical threshold ranges, and granular alignment evaluation of timestamp accuracy level and delivery requirement data. This quantifies the degree of semantic deviation in each dimension and generates a multi-dimensional semantic deviation vector.

[0081] The system receives a set of structured metadata to be compared and a set of standardized semantic contracts, and initiates a multi-dimensional semantic consistency verification method. For field name strings, the edit distance method is used to calculate the character difference between the source field name and the standard attribute name, generating a name matching score. For the original unit of measurement identifier, its dimensional exponent vector is parsed and a dot product operation is performed with the unit constraint dimensions in the snapshot to verify dimensional compatibility. If the dimensions are consistent but the units are different, it is marked as a unit conversion requirement. For real-time collected numerical ranges, boundary inclusion detection is performed to determine whether the values ​​fall within the threshold range specified by business constraints. The numerical deviation is calculated to quantify the degree of data anomaly. For timestamp accuracy levels, the time granularity of the source data is compared with the granularity standard of the delivery requirement data to evaluate time alignment. If the source data accuracy is lower than required, it is marked as insufficient accuracy. Combining the verification results of the above four dimensions, a four-dimensional semantic deviation vector is constructed. Each component of this vector corresponds to a deviation value in one dimension; zero indicates a complete match, and non-zero values ​​indicate a deviation. By using a multidimensional semantic consistency verification method, the structured meta-information from the previous step is transformed into a multidimensional semantic deviation vector that quantifies the degree of semantic deviation in each dimension, thus realizing the transformation from qualitative description to quantitative evaluation and providing data support for the accurate screening of subsequent difference fields.

[0082] For example, taking cleanroom differential pressure sensor data as an example, the source field name is "Press_kPa", and the standard attribute name is "PressureDifference". The calculated edit distance score is 0.6, which is below the threshold of 0.8, and is marked as name deviation. The source unit is kPa, and the standard unit is Pa, both with units of ML. -1 T -2 The measurements are dimensionally compatible, but unit conversion is required. The collected value is 15.5 kPa, which, when converted to 15500 Pa, exceeds the business constraint of 10-50 Pa, indicating a high deviation. The timestamp accuracy is in the second range, while the delivery requirement is in the millisecond range, thus marking it as insufficient precision. The generated four-dimensional semantic deviation vector is [0.4, 0, 1.0, 0.5], where the deviation values ​​for name, unit, value, and time dimensions are 0.4, 0, 1.0, and 0.5, respectively. This vector accurately reflects the multi-dimensional semantic deviation of the data, guiding subsequent micro-instruction generation.

[0083] S3.4: Based on the non-zero deviation components in the multidimensional semantic deviation vector, perform difference feature marking processing on data items with semantic conflicts, and filter out abnormal data items such as field name string matching failure, inconsistent original unit of measurement identification dimensions, real-time collected value range exceeding the value threshold range, or timestamp accuracy level lower than the delivery requirement data standard, so as to construct a set of difference fields containing specific deviation types and deviation values.

[0084] Receive a multidimensional semantic deviation vector, iterate through each dimension component of the vector, and perform non-zero value determination logic. Identify the component indices with non-zero values, map them to the corresponding data field identifiers, and mark them as discrepancies to be processed.

[0085] For cases where string matching of field names fails, the source data field names are extracted and compared with the standard field names in the standardized semantic contract. The edit distance or similarity score between the two is calculated. If the score is lower than a preset threshold, it is determined to be a naming deviation, the deviation type is recorded as 'NameMismatch', and the source field name is retained as the basis for correction.

[0086] To address situations where the original units of measurement have inconsistent dimensions, the unit symbols in the source data are analyzed and compared with the standard unit symbols in the snapshot. The physical dimension exponent vectors of the two are compared using dimensional analysis methods. If the dimensions are incompatible, it is marked as 'UnitIncompatible'; if the dimensions are compatible but the units are different, it is marked as 'UnitDeviation', and the required conversion factor is recorded.

[0087] For cases where the real-time collected numerical range exceeds the numerical threshold, the measured values ​​of the source data are compared with the upper and lower thresholds defined in the snapshot. Inequality judgment logic is constructed as follows: or

[0088] in These are measured values. and These are the minimum and maximum thresholds, respectively. When either condition is met, it is marked as 'ValueOutOfRange', and the extent of exceeding the limit is recorded.

[0089] For data where the timestamp precision level is lower than the delivery requirement standard, parse the effective number of bits or resolution of the source data timestamp. Compare it with the minimum time granularity specified in the snapshot. If the source data precision is insufficient, mark it as 'PrecisionInsufficient' and record the required interpolation or alignment strategy.

[0090] Integrate all marked outlier data items to construct a structured set of difference fields. Each set element contains a field ID, deviation type code, original value, expected value, and deviation quantification parameter.

[0091] Through the above item-by-item comparison and feature labeling process, the multidimensional semantic deviation vector of the previous step is transformed into a set of difference fields containing specific deviation types and values, realizing the accurate location and structured description of semantic conflicts in multi-source heterogeneous data, and providing clear input for the subsequent generation of micro-execution instructions.

[0092] For example, a differential pressure sensor reports data with the field name 'diff_press_kpa', a value of 0.025, in kPa, and a timestamp accuracy in the second range. The corresponding BIM component snapshot requires a field name of 'PressureDifference', in Pa, with a threshold range of 15-50 Pa and a time accuracy in the millisecond range. The system identifies a mismatch in field names, marking it 'NameMismatch'; it identifies that the unit kPa is compatible with Pa but the value needs conversion, marking it 'UnitDeviation' with a conversion factor of 1000; it calculates 0.025 kPa = 25 Pa, which falls within the 15-50 Pa range, so there is no 'ValueOutOfRange' error; and it identifies that the timestamp accuracy in the second range is lower than the millisecond requirement, marking it 'PrecisionInsufficient'. Finally, a set of difference fields containing these three types of deviation information is generated, directly driving the generation of subsequent micro-instructions.

[0093] S3.5: Integrate the set of difference fields with the corresponding multidimensional semantic deviation vector, and use a difference encoding protocol to encapsulate and generate a difference-aware result object containing a list of difference fields, a deviation type identifier, and correction suggestion parameters, so as to complete the semantic conversion closed loop from the original data stream to the executable difference instruction set.

[0094] Step S4: Process the difference perception result to obtain a set of field path redirection instructions, a unit conversion coefficient field, and an enumeration value mapping table, and encapsulate the set of field path redirection instructions, the unit conversion coefficient field, and the enumeration value mapping table into a semantic mapping instruction set.

[0095] This step addresses the issue of low format conversion efficiency in the fusion of multi-source heterogeneous data. By constructing an automated mapping mechanism between difference-aware results and micro-execution instructions, it transforms identified semantic deviations into a set of executable, lightweight operations. It plays a crucial role in the overall technical solution, both bridging the gap left by the precise location of data differences in the preceding steps and providing standardized, low-latency instruction inputs for the subsequent atomic transformation of edge nodes, ensuring efficient incremental fusion without sacrificing data consistency.

[0096] S4.1: Based on the field names in the difference perception results, perform topology matching processing on the source data field paths and the standard attribute paths in the target component-level semantic state snapshot to generate a set of field path redirection instructions containing source path pointers and target path pointers, thereby establishing the physical mapping relationship of data flow.

[0097] The system receives field names from the difference-aware results and extracts the original path strings of the source data fields and the standard attribute path identifiers from the target component-level semantic state snapshot. It performs hierarchical parsing of the source data paths, constructing a source path topology tree structure containing node names and index positions. It then performs industrial base class schema mapping analysis on the standard attribute paths, generating a target path topology tree structure containing class definitions and attribute set references. A bidirectional breadth-first search method is initiated to calculate the edit distance and semantic similarity matrix between the source and target path topologies. Based on the maximum similarity principle, a unique mapping relationship is established from source data nodes to target industrial base class attributes, generating a redirection instruction tuple containing source and target pointer offsets. Through topology matching, the non-standardized source field paths are transformed into precise physical mapping instructions pointing to BIM component attributes, achieving zero-ambiguity routing and positioning of heterogeneous data sources.

[0098] S4.2: Using the source data unit type determined in the field path redirection instruction set and the standard unit type in the target component-level semantic state snapshot, the preset unit conversion coefficient lookup table is called to perform dimensional consistency calculation processing to generate a unit conversion coefficient field containing multiplication factor values, thereby eliminating dimensional heterogeneity between multiple source data.

[0099] The system receives the dimension inconsistency field from the difference perception results and extracts the source data unit identifier and the standard unit constraints from the target component-level semantic state snapshot. It retrieves a pre-set unit conversion factor lookup table, stored in read-only memory, indexing linear transformation parameters using source-target unit pairs as keys. It determines the physical dimension compatibility between the source and target units; if the dimensions do not match, processing terminates and an error is reported to ensure consistency in physical meaning. The system then calls the dimension consistency calculation module, determining the conversion logic based on the International System of Units (SI) derivation rules. For linear proportional relationships, the unit conversion factor is calculated using the following formula:

[0100] in, For unit conversion factors, The benchmark value for the target standard unit. The source data serves as the unit reference value. For nonlinear units involving offsets, such as those from Celsius to Kelvin, multiplication factors and addition constants are extracted to construct an affine transformation parameter set. The calculated multiplication factor values ​​are encapsulated in floating-point format, retaining at least 6 significant digits to maintain the data integrity of high-precision sensors. A unit conversion coefficient field containing the multiplication factor is generated, marking its associated source field path and target attribute ID. Through the above processing method, the dimensional deviations identified in the previous step are transformed into accurate mathematical correction parameters, achieving standardization and unification of physical dimensions among multi-source data and eliminating data fusion errors caused by unit heterogeneity.

[0101] S4.3: Based on the business constraint rules in the difference perception results, perform key-value pair mapping processing on the discrete state values ​​of the source data and the enumerated state list in the target component-level semantic state snapshot to generate an enumerated value mapping table containing the correspondence between the source state code and the target state code, thereby solving the problem of inconsistent business state descriptions.

[0102] The system receives the difference-aware results and extracts the discrete state value strings from the source data and the standard enumerated state set defined in the target component-level semantic state snapshot. It performs Unicode standardization preprocessing on the source data state values ​​to eliminate matching interference caused by full-width / half-width characters, invisible control characters, and leading / tailing whitespace, forming a standardized source state identifier. Based on the edit distance method, it calculates the string similarity between the standardized source state identifier and each candidate option in the standard enumerated set, constructing an initial fuzzy matching matrix. It introduces a context weight factor from business constraint rules, assigning higher matching priority to state words in specific construction processes (such as "debugging phase"), thus correcting the initial similarity score. A dynamic threshold judgment mechanism is set: when the highest similarity score exceeds a preset confidence interval, a unique mapping relationship is established; if there is ambiguity conflict, manual intervention is triggered or the system defaults to an "unknown" state code. The confirmed source state code and target standard state code are encapsulated into a key-value pair structure, generating a compact binary enumerated value mapping table. Through the above processing method, the difference perception results of the previous step are transformed into an executable discrete state transition instruction set, realizing accurate alignment and unambiguous fusion of multi-source heterogeneous data at the business semantic level.

[0103] For example, the status value "Closed" reported in the cleanroom airtight door construction log is mapped to the standard enumeration set {"closed", "open", "maintenance"} defined in the BIM snapshot. After Unicode cleansing, "Closed" remains unchanged. Its semantic relevance to "closed" is calculated, and a weighting coefficient of 1.2 is assigned, considering the current process is "airtightness test". Assuming a basic similarity of 0.85, the weighted score is 1.02, exceeding the threshold of 0.9, thus confirming a valid mapping. A mapping entry is generated: the source code "Closed" points to the target code "closed". This process takes less than 2ms, ensuring semantic consistency of status data during integration and avoiding data loss or misjudgment due to description differences.

[0104] S4.4: Integrates the set of field path redirection instructions, unit conversion coefficient fields, and enumeration value mapping tables, and uses a fixed-length binary encoding method for serialization and encapsulation to generate a semantic mapping instruction set with strictly limited length, thereby meeting the stringent requirements of edge computing nodes for instruction loading speed and memory usage.

[0105] like Figure 3As shown, step S5 involves loading the semantic mapping instruction set onto an edge computing node for execution, performing atomic transformation processing on the multi-source data stream, generating standardized fused data, and completing the integration of digital delivery information for engineering projects.

[0106] This step is responsible for deploying the semantic mapping instruction set to edge computing nodes, using an atomic execution mechanism to transform multi-source data streams in real time, and generating standardized fused data that conforms to the constraints of component-level semantic state snapshots. It is the core execution link for achieving low-latency, high-consistency data integration and dynamic updates.

[0107] S5.1: Obtain the semantic mapping instruction set and multi-source data stream, and perform loading verification processing on the semantic mapping instruction set based on the memory isolation mechanism of the edge computing node to generate a verified micro-execution instruction sequence containing field path redirection instructions, unit conversion coefficients and enumeration value mapping tables.

[0108] Receive the semantic mapping instruction set generated in step S4, which includes field path redirection pointers, unit conversion factors, and an enumerated value mapping table. Simultaneously, access the multi-source data streams to be processed and extract their original payload content.

[0109] Initialize a memory isolation region inside the edge computing node and allocate independent stack space for instruction loading and execution to prevent memory overflow or illegal access during instruction execution from interfering with the operation of the main system.

[0110] Integrity checks are performed on the loaded instruction set. The hash value of the instruction binary stream is calculated and compared with the checksum in the metadata header to ensure that the instructions have not been tampered with or damaged during transmission.

[0111] Parse the field path redirection pointers in the instruction set, verify the existence of the source path in the original data JSON structure, and the validity of the target path in the BIM component industrial basic class attribute set, and establish a two-way mapping index.

[0112] Read the unit conversion coefficient field, check its numerical validity and dimensional compatibility, and exclude abnormal instruction items with a divisor of zero or a non-numerical type to ensure the mathematical stability of subsequent linear transformations.

[0113] Traverse the enumeration value mapping table to confirm the one-to-one correspondence between the source status code and the target status code, mark the discrete values ​​of missing mappings as anomalies to be processed, and generate a complete and verified micro-execution instruction sequence.

[0114] Through the aforementioned memory isolation loading and multi-verification mechanism, the original instruction set generated in the previous step is transformed into a secure, legal, and immediately executable verified micro-execution instruction sequence, achieving high reliability and low latency startup for edge-side data conversion.

[0115] S5.2: Receive the original data packet from the multi-source data stream, and use the field path redirection instruction in the verified micro-execution instruction sequence to perform atomic mapping processing on the key-value structure of the original data packet to generate an intermediate structured data object carrying a standard attribute set identifier.

[0116] Receives the micro-execution instruction sequence verified by S5.1 and the raw JSON or XML messages from the multi-source data stream. Parses the key-value pair structure of the raw message, extracting the source field path string to be mapped and its corresponding raw value or status value. Traverses the field path redirection instruction set in the verified micro-execution instruction sequence, reading the source path pointer and target path pointer contained in each instruction. Uses a depth-first search method to locate the specific data node pointed to by the source path pointer in the nested structure of the raw message. If the source data node exists and the type matches, reads the current value of the node as the payload to be converted. Based on the hierarchical structure of the industrial basic class attribute set defined by the target path pointer, constructs a temporary intermediate structured data object in memory. Writes the payload to be converted to the storage location of the corresponding target path in the intermediate object and marks that the field has completed physical mapping. For redundant fields in the raw message that do not appear in the redirection instruction set, performs discard processing to reduce memory usage. Checks whether all required standard attribute set identifiers in the intermediate structured data object have been assigned values. If any unmapped key semantic fields exist, the object is marked as incomplete and an exception handling process is triggered. Through atomic field redirection, the original heterogeneous message from the previous step is transformed into an intermediate structured data object carrying a standard industrial basic class attribute set identifier, achieving standardized alignment of the data's physical structure and providing a unified data carrier for subsequent dimensional unification and semantic normalization.

[0117] For example, in a JSON data packet {"dev_id": "SENS_01", "val": 25.5, "unit": "Pa"} reported by a cleanroom differential pressure sensor, the micro-execution instruction specifies the source path as "$.val" and the target path as "Industrial Basic Class Sensor.Pset_SensorCommon.Value". The system locates the "val" field in the JSON, obtains the value 25.5, creates an intermediate object conforming to the Industrial Basic Class standard in memory, writes 25.5 into the Value property, and binds it to the Pset_SensorCommon identifier. The non-standard "dev_id" field in the original packet is discarded because there is no mapping instruction. The final generated intermediate object contains only a standardized attribute set structure and values, eliminating interference from private protocol fields and significantly improving the structural consistency of subsequent data processing.

[0118] S5.3: Extract the numerical fields and their original units of measurement from the intermediate structured data object, and perform a linear scaling transformation operation based on the corresponding unit conversion coefficients in the verified micro-execution instruction sequence to generate a normalized physical quantity data set that conforms to the International System of Units (SI).

[0119] S5.4: Traverse the state enumeration field in the normalized physical quantity data set, and perform discrete value matching and replacement operation by referring to the enumeration value mapping table embedded in the verified micro-execution instruction sequence to generate a normalized business state parameter list with completely consistent semantic values.

[0120] The system receives intermediate structured data objects that have undergone unit normalization and parses the discrete field identifiers containing business status descriptions. It extracts the corresponding enumeration value mapping table from the verified micro-execution instruction sequence. This mapping table defines the key-value pair relationship between non-standard status codes from the source system and standard enumeration values ​​in the BIM component-level semantic status snapshot. It iterates through each status enumeration field in the dataset to be processed, reading its current original discrete value. It retrieves the source status code key-value that matches the original discrete value from the enumeration value mapping table. If the search is successful, it obtains the corresponding target standard status code from the mapping table, performs an atomic replacement operation, and updates the original discrete value to the standard enumeration value. If the search fails, it marks the field as a semantically abnormal state and records the original value for manual review. All status fields that have been replaced or marked are reassembled to form a standardized business status parameter list. Through discrete value matching and replacement processing, the numerical normalized data from the previous step is transformed into a standardized business status parameter list with completely consistent semantic values, achieving consistent fusion of multi-source heterogeneous data at the business logic layer.

[0121] For example, for a cleanroom airtight door component, the source construction log reports the status as "sealed," and the BIM semantic snapshot defines the standard enumeration as "sealed." The enumeration value mapping table contains key-value pairs {"sealed": "sealed", "open": "open"}. The system reads the field value "sealed," finds a match in the mapping table, and replaces it with "sealed." If the reported value is "closing," and there is no corresponding key in the mapping table, the system marks the field as abnormal. The final output contains a normalized parameter list with the standard status "sealed," ensuring the consistency of status semantics during subsequent delivery audits.

[0122] S5.5: Integrate the standardized business status parameter list with the delivery requirement data in the latest component-level semantic status snapshot, perform integrity verification and format encapsulation processing to generate standardized fused data that strictly meets the constraints of the component-level semantic status snapshot.

[0123] The system receives a standardized business status parameter list and retrieves delivery requirement data from the latest component-level semantic status snapshot. It parses the required attribute set IDs and document reference path constraints within the delivery requirement data. It iterates through the standardized business status parameter list, extracting currently mapped attribute key-value pairs. It performs an intersection operation between the extracted attribute key-value pairs and the required attribute set IDs to identify missing key delivery attribute fields. If missing fields exist, it performs completion processing based on the default value strategy or null value marking rules preset in the snapshot, generating a structured data temporary storage area that passes integrity verification. It performs industrial basic class standard format serialization encapsulation on the structured data temporary storage area, mapping attribute data to the corresponding industrial basic class ObjectDefinition instance. During serialization, it injects a unique component identifier and snapshot version hash as metadata header information. It uses JSON-LD or XML Schema to define and verify the legality of the encapsulated data structure. Through integrity verification and standardized format encapsulation, it transforms discrete business parameters into structured fused data conforming to BIM delivery standards, achieving seamless conversion from multi-source data to standardized deliverables.

[0124] For example, for a cleanroom HEPA filter (FFU) component, the standardized parameter list includes a differential pressure value of 250 Pa and a completed status. Delivery requirements stipulate that the FilterEfficiency attribute must be included. The system detects this missing attribute and fills it in according to the snapshot default value in ISO 14644-1 Class 5. Subsequently, the differential pressure, status, and efficiency attributes are encapsulated into an instance of the industrial base class PropertySet, associated with an industrial base class Filter object, and a COBie-compliant JSON delivery data package is generated, ensuring the data is directly usable for as-built model review.

[0125] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this invention.

[0126] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains. The terms “first,” “second,” “third,” and similar terms used in this patent application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an” or “a” and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms “comprising” or “including” and similar terms mean that the elements or objects preceding “comprising” or “including” encompass the elements or objects listed following “comprising” or “including” and their equivalents, and do not exclude other elements or objects. The “multiple” mentioned in the embodiments of this application refers to two or more. A and / or B indicate three possibilities: A; B; and A and B.

[0127] The above description is merely an exemplary embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and such modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for integrating digital delivery information for cleanroom renovation projects, characterized in that, Specifically, it includes: S1: Obtain the structural definition information, business constraint rules, and delivery requirement data of key components in the BIM model of the cleanroom decoration project; generate component-level semantic state snapshots based on the structural definition information, business constraint rules, and delivery requirement data; and combine multiple component-level semantic state snapshots arranged in time series to construct a component-level snapshot chain. S2: Receive multi-source data streams from sensors, construction logs, or design drawings; construct a composite index key based on the multi-source data streams; and use the composite index key to locate and obtain the latest component-level semantic state snapshot in the component-level snapshot chain. S3: Compare the multi-source data stream with the latest component-level semantic state snapshot, identify the set of discrepancies in field name, unit, value range or time precision, and generate a difference perception result based on the set of discrepancies. S4: Process the difference perception result to obtain a set of field path redirection instructions, a unit conversion coefficient field, and an enumeration value mapping table, and encapsulate the set of field path redirection instructions, the unit conversion coefficient field, and the enumeration value mapping table into a semantic mapping instruction set; S5: Load the semantic mapping instruction set to the edge computing node for execution, perform atomic transformation processing on the multi-source data stream, generate standardized fused data, and complete the integration of digital delivery information for engineering projects.

2. The method for integrating digital delivery information for cleanroom renovation projects according to claim 1, characterized in that, S1 specifically includes: Based on the geometric topology in the BIM model of the cleanroom decoration project, semantic entity recognition processing is performed on key components such as the installation points of high-efficiency filters, airtight doors and differential pressure sensors to obtain structural definition information including attribute set IDs. Using the attribute set ID in the structure definition information, the industry standard library and project delivery standards are associated, and business rule instantiation processing is performed on each key component to generate business constraint rules; Based on the mapping relationship between the business constraint rules and the engineering acceptance process, the document reference path and deliverable association logic of key components are encapsulated to form delivery requirement data containing read-only attribute binding requirements. By integrating the aforementioned structural definition information, business constraint rules, and delivery requirement data, a component-level semantic state snapshot containing a unique version hash value and a high-precision timestamp is generated using a serialization method. Based on the timestamp sequence of the component-level semantic state snapshots, chain linking is performed on multiple component-level semantic state snapshots of the same key component at different engineering stages to construct a component-level snapshot chain.

3. The method for integrating digital delivery information for cleanroom renovation projects according to claim 2, characterized in that, The structure definition information also includes an industrial basic class type identifier.

4. The method for integrating digital delivery information for cleanroom renovation projects according to claim 2, characterized in that, The business constraint rules include unit constraints, numerical threshold ranges, and enumerated state lists.

5. The method for integrating digital delivery information for cleanroom renovation projects according to claim 2, characterized in that, S2 specifically includes: Perform protocol parsing and payload extraction processing on multi-source data streams to separate metadata header information; The metadata header information is processed using a multidimensional key-value concatenation method to generate an engineering context identifier, in order to construct a composite index key that can uniquely point to a specific BIM component instance; Using the composite index key as a query condition, hash addressing and version retrieval processing are performed in the index database to lock the set of component-level semantic state snapshots of all historical versions of the target key component in the current time window; The latest component-level semantic state snapshot is obtained by sorting and performing operations on the set of component-level semantic state snapshots.

6. The method for integrating digital delivery information for cleanroom renovation projects according to claim 5, characterized in that, The index database stores a complete chain of component-level snapshots.

7. The method for integrating digital delivery information for cleanroom renovation projects according to claim 5, characterized in that, The process of sorting and performing operations on the set of component-level semantic state snapshots to obtain the latest component-level semantic state snapshot includes: Based on the high-precision timestamp and unique version hash value carried by each snapshot in the component-level semantic state snapshot set, the time-series sorting and latest state determination logic operation is performed to filter out the latest component-level semantic state snapshot from multiple historical versions.

8. The method for integrating digital delivery information for cleanroom renovation projects according to claim 1, characterized in that, S4 specifically includes: Based on the field names in the difference perception results, a set of field path redirection instructions is generated. Using the source data unit type determined in the set of field path redirection instructions and the standard unit type in the target component-level semantic state snapshot, a dimensional consistency calculation is performed to generate a unit conversion coefficient field containing multiplication factor values. Based on the business constraint rules in the difference perception results, key-value pair mapping is performed on the discrete state values ​​of the source data and the enumerated state list in the target component-level semantic state snapshot to generate an enumerated value mapping table containing the correspondence between the source state code and the target state code. The system integrates the set of field path redirection instructions, unit conversion coefficient fields, and enumeration value mapping tables, and uses a fixed-length binary encoding method for serialization and encapsulation to generate a semantic mapping instruction set with strictly limited length.

9. The method for integrating digital delivery information for cleanroom renovation projects according to claim 8, characterized in that, The set of field path redirection instructions generated based on the field names in the difference-aware results includes: Based on the field names in the difference perception results, topology matching is performed on the source data field paths and the standard attribute paths in the target component-level semantic state snapshot to generate a set of field path redirection instructions.

10. The method for integrating digital delivery information for cleanroom renovation projects according to claim 8, characterized in that, The set of field path redirection instructions includes source path pointers and target path pointers.