Progress management system for engineering construction project digitization

Abstract

The invention relates to a progress management system for engineering construction project digitization. The system comprises a model uploading and lightweight module, a browser end rendering module, an annotation management module, a risk transformation and task management module and a data archiving and synchronizing module. Three core functions of zero client review, real-time collaborative annotation and risk intelligent early warning of a design model are realized through cooperative work among the modules in the system and a lightweight engine and structured data binding technology.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of engineering project management, in particular, the present application relates to a progress management system for digital engineering construction project. BACKGROUND

[0002] In the prior art, although energy engineering has achieved some results in digital transformation, it has not yet formed a digital collaborative and visual management and control system covering the whole process (design, procurement, construction, delivery). Therefore, it is necessary to establish a whole-process digital project management system to strictly coordinate the business of PMC, supervision and detection on a complete, digital, visual and intelligent platform, so as to realize the goals of improving work quality, work efficiency, one-key delivery of completion data and empowering business through digital platform.

[0003] However, the engineering progress management in the current digital project management system has the following pain points: traditional progress detection relies on manual reporting, data is lagging and prone to distortion; three-dimensional model and progress data are disconnected, and the visualization degree is low; PMC, supervision, detection and other business links lack a unified digital platform, resulting in information silos and collaboration barriers. Relying on professional CAD software (such as AutoCAD / SolidWorks), a 3GB+ client needs to be installed, mobile support is missing, on-site engineers cannot review in real time, cross-platform collaboration requires format conversion, and the data loss rate is as high as 18%. Comments are delivered through email / paper, with an average response cycle of 72 hours; design change applications are disconnected from models, version management is chaotic, design risk identification is lagging, and quality problems are passed on to the construction phase. Data silos: design, supervision, construction data are separated, welding detection and other on-site data cannot be associated with models, domestic adaptation is insufficient, and there are information security risks.

[0004] For example, in the prior art CN116611792A, a project bidding processing unit for bidding agent is included, wherein the project bidding unit includes implementing project funds, submitting the project to the relevant administrative department for approval; starting the bidding work; reminding to submit the bidding document to the bidding site before the deadline of submitting the bidding document according to the requirements of the bidding document; uploading and reviewing the bidding document, in the present application, the total planned construction progress is input into the cloud database to generate a model, then the number of daily construction personnel and daily construction equipment is input to generate an actual daily construction progress model, then monthly or weekly summaries are made and compared with the planned construction progress of each month or week, so that project personnel can timely understand whether the construction progress is completed within the quantitative time, facilitating construction project management. However, CN116611792A mainly focuses on the progress management of engineering projects, does not involve structured linkage and whole-cycle management of the whole process, and mainly focuses on the function of a single module, lacking system integration and collaboration function.

[0005] In the prior art CN110163509A, a project full information plan module is used to record project annual plan information and monthly plan information reported by each department; a project first-level node plan module is used to record project first-level node plan information, which at least includes a plan schedule, and the plan schedule includes a planned task item and a corresponding planned execution start and end time; a project actual progress reporting module is used to receive project actual completion progress reported by a business department related to engineering construction; a project progress management module is used to summarize and record the project actual completion progress reported by the project actual progress reporting module, and compare the actual completion progress with the plan schedule to record a comparison result; and a project plan ledger module is used to generate project full plan ledger information. However, CN110163509A still stays in the "form + email" type progress management: manual reporting causes data lag and distortion; three-dimensional models are separated from progress, quality, and detection data, forming an information island; and GB-level clients are relied on without mobile terminals, and real-time review cannot be performed on site.

[0006] Therefore, there is an urgent need in the prior art for a lightweight three-dimensional visual progress management and control system without the need to install a client and supporting cross-terminal real-time review, which can dynamically bind three-dimensional models with progress, quality, and detection data, realize a comment-risk-task-archiving closed-loop driving, solve the problems of manual reporting lag, version management confusion, information island, and lack of localization security, and meet the needs of large-scale engineering such as petrochemical engineering and long-distance pipeline for millisecond-level response, millimeter-level positioning, responsibility to person, and auditable digital progress management data. SUMMARY

[0007] The technical problem to be solved by the present application is to provide an engineering construction project digital progress management system, which aims to solve at least one of the above technical problems.

[0008] In a first aspect, the technical solution of the present application to solve the above technical problems is as follows: an engineering construction project digital progress management system, comprising: a model uploading and lightweight module, configured to receive a three-dimensional model file of an engineering to be audited uploaded by a user, perform format analysis and face simplification on the three-dimensional model file, and obtain a lightweight model; a browser-side rendering module, configured to dynamically load the lightweight model with zero plug-ins, and provide an interactive function for the user to design and review the lightweight model online across terminals, the interactive function including interactive functions of rotation, sectioning, measurement, and comment positioning; The batch note management module is used for binding the review opinions input by the user on the lightweight model with the model space coordinates to generate a structured batch note, and tracking, synchronizing and archiving the whole life cycle state of the structured batch note, so as to realize closed-loop management of design problems in the three-dimensional model file. The risk conversion and task management module is used for keyword extraction and risk library matching of the structured batch note, automatic generation of corresponding risk items and rectification tasks, real-time pushing of the rectification tasks to the responsible person and time limit control and closed-loop verification of the rectification tasks according to the state machine. The data archiving and synchronizing module is used for synchronizing the processing data to the risk library, the task system and the completion database, so as to ensure the closed-loop management, and the processing data is the structured batch note, the risk, the task, the change and the completion data generated in the engineering closed-loop design review process.

[0009] The beneficial effects of the present application are that: through the integrated architecture of'model lightweight + browser zero plug-in rendering + batch note-risk-task closed-loop driving + multi-terminal data synchronization', the GB-level model is compressed to the size that can be loaded in seconds on the webpage, cross-terminal online review is realized; the review opinions are bound to the model with millimeter-level precision and structured batch notes are automatically generated, after keyword extraction and risk library matching, the graded rectification tasks are generated and pushed to the responsible person in time, the task state is synchronized in real time; the processing data can be synchronized to the risk library, the task system and the completion database after being collected once, the manual reporting lag, version disconnection and information island are eliminated, the digital progress closed loop with auditability, traceability and person responsibility is formed, and the progress control efficiency and data integrity are significantly improved.

[0010] On the basis of the above technical solutions, the present application can also be improved as follows.

[0011] Further, the above model uploading and lightweight module is specifically used for: determining the original file format of the three-dimensional model file; converting the three-dimensional model file into geometric data of a unified data structure according to the original file format; performing triangle facet simplification processing on the geometric data to obtain a simplified triangular mesh model which retains key engineering features and has reduced data volume; performing compression processing on the simplified triangular mesh model to obtain a lightweight model.

[0012] Further, when the above model uploading and lightweight module converts the three-dimensional model file into geometric data of a unified data structure according to the original file format, it is specifically used for: if the original file format is a native format, the three-dimensional model file is directly subjected to geometric topology analysis to obtain geometric data of a unified data structure; If the original file format is an intermediate format, the three-dimensional model file is standardized by a converter to obtain geometric data with a unified data structure.

[0013] Further, the model uploading and lightweight module is used for: The improved QEM algorithm is used to perform secondary error measurement on the simplified triangular mesh model to form a re-compressed mesh. The BasisUniversal super compression algorithm is used to convert the model texture corresponding to the simplified triangular mesh model from the original format to the target format to obtain texture data that can be directly decompressed by the GPU. Non-geometric data in the simplified triangular mesh model is removed to obtain engineering metadata, which is encoded according to the JSON-LD standard to obtain structured metadata. The lightweight model is generated according to the re-compressed mesh, texture data, and structured metadata.

[0014] Further, the browser-side rendering module includes: The block transmission unit is used to divide the lightweight model into multiple blocks according to spatial regions, and to implement the download of the multiple blocks by HTTPRange request. The local cache unit is used to cache the downloaded blocks by ServiceWorker and to store the model metadata corresponding to all blocks in IndexedDB, which is read locally during secondary loading.

[0015] Further, the block transmission unit is also used to: LOD0-3 four-level meshes are configured for different components in each block of the lightweight model, and the meshes of the corresponding level are dynamically switched for download within the same block according to camera distance and device performance, wherein the accuracy of LOD0 to LOD3 decreases in turn.

[0016] Further, the browser-side rendering module also includes: The dynamic sectioning unit is used to perform real-time sectioning of the lightweight model on the GPU side by a WebGL fragment shader. The avatar roaming unit is used to provide WASD key control in first-person camera mode and to realize collision reduction between the first-person camera and the model surface of the lightweight model by ray detection. The component display and measurement unit is used to hide or display corresponding components based on the model tree Group level according to filtering conditions, and to perform precision measurement by a spatial two-point distance algorithm.

[0017] Further, the browser-side rendering module also includes: The device capability detection subunit is configured to determine the GPU performance of the device in the loading stage through WebGLRenderer.capabilities, so as to automatically adjust the rendering parameters according to the GPU performance of the device; The touch interaction adaptation subunit is configured to generate the physical view transformation parameters corresponding to different operation gestures based on the mobile terminal, and set the operation sensitivity coefficient according to the screen size, so as to uniformly map the physical view transformation parameters to the model view transformation parameters based on the operation sensitivity coefficient; The performance monitoring and dynamic adjustment subunit is configured to automatically reduce the rendering quality parameter in the rendering parameters when the frame rate is lower than the preset threshold for the first set duration.

[0018] Further, the above-mentioned comment management module specifically comprises: The millimeter-level positioning subunit is configured to convert the two-dimensional pixel coordinates corresponding to the review opinions input by the user on the lightweight model into model space coordinates, and correct the converted three-dimensional coordinates by the vertex offset amount before and after the pre-computed model simplification; The structured comment subunit is configured to bind the business attributes including the model ID, the component ID, the professional type, the priority, the attachment ID and the referenced standard clause for each block, and store them in the JSON format and transmit them in the Protobuf protocol; The life cycle subunit is configured to sequentially perform the three verifications of the responsibility person role matching claim, the rectification evidence submission and the supervision audit on the comment state conversion request according to the conversion rules preset in each stage of the whole life cycle of the to-be-audited project, the comment state conversion request being an electronic instruction for pushing the specified comment from the current state to the next state in the whole life cycle; after the three verifications are passed, the specified comment is pushed from the current state to the next state in the whole life cycle, and the corresponding SHA-256 chain fingerprint log is generated; The collaboration and report subunit is configured to merge the concurrent comments of multiple users by using the operation conversion algorithm, and automatically generate the vector PDF review report with the 3D position snapshot, the standard clause and the electronic seal according to the professional, the problem type and the processing state.

[0019] Further, the above-mentioned risk transformation and task management module specifically comprises: The risk intelligent identification subunit is configured to perform the named entity recognition and the relationship extraction on the structured comment by using the pre-trained keyword extraction model, obtain the triplets and automatically match them with the HSE risk library, and output the P0-P2 classification results according to the matching results; The intelligent task generation subunit is configured to automatically generate the rectification task containing the task name, the rectification requirement, the model snapshot, the standard clause and the response time limit according to the P0-P2 classification results, wherein the time limit corresponding to the rectification task of different levels is different; A person-in-charge matching subunit is configured to push the rectification task to the person-in-charge. A closed-loop verification subunit is configured to perform model version comparison, OCR evidence identification and supervision electronic signature acceptance in sequence after the rectification task is submitted, and complete the closed-loop verification of the rectification task.

[0020] Additional aspects and advantages of the application will be described in the following description, will become apparent from the following description, or will be learned by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments of the present application will be briefly introduced.

[0022] Figure 1 A schematic diagram of a progress management system for digital engineering construction projects is provided for an embodiment of the present application; Figure 2 A business process diagram is provided for an embodiment of the present application; Figure 3 A processing flow schematic diagram of model uploading and lightweight module is provided for an embodiment of the present application; Figure 4 A model loading schematic diagram is provided for an embodiment of the present application; Figure 5 An operation interface schematic diagram is provided for an embodiment of the present application; Figure 6 An interaction processing logic schematic diagram is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0023] The principles and features of the present application are described below, and the examples are only used to explain the present application, and are not used to limit the scope of the present application.

[0024] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems will be described in detail below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes can not be described again in some embodiments. The embodiments of the present application will be described below with reference to the drawings.

[0025] The scheme provided by the embodiments of the present application can be applied to any application scenario that needs to manage the progress of digital engineering construction projects.

[0026] The embodiments of the present application provide a possible implementation manner, as Figure 1 shown, a schematic diagram of a progress management system for digital engineering construction projects is provided, and the schematic diagram is combined with Figure 2The illustrated business process diagram illustrates the method provided by the embodiment of the application, as shown in Figure 1 The system can include: A model uploading and lightweight module, configured to receive a three-dimensional model file of an engineering to be audited uploaded by a user, perform format analysis and facet simplification on the three-dimensional model file, and obtain a lightweight model; A browser-side rendering module, configured to dynamically load the lightweight model with zero plug-ins, and provide an interactive function for the user to design and review the lightweight model online across terminals, the interactive function including interactive functions of rotation, sectioning, measurement and comment positioning; A comment management module, configured to bind a review opinion input by the user on the lightweight model to a model space coordinate to generate a structured comment, and track, synchronize and archive a whole life cycle state of the structured comment, so as to realize closed-loop management of a design problem in the three-dimensional model file; A risk conversion and task management module, configured to extract a keyword from the structured comment and match the keyword with a risk library, automatically generate a risk item and a rectification task corresponding to a corresponding level, push the rectification task to a person in charge in real time, and complete time control and closed-loop verification of the rectification task according to a state machine; A data archiving and synchronizing module, configured to synchronize processing data to the risk library, a task system and a completion database, so as to ensure closed-loop management, and the processing data is structured comment, risk, task, change and completion data generated in a closed-loop design review process of the engineering to be audited.

[0027] The scheme of the application compresses a GB-level model to a size that can be loaded in seconds on a webpage through an integrated architecture of “model lightweight + browser zero plug-in rendering + comment-risk-task closed-loop driving + multi-terminal data synchronization”, realizes online review across terminals, binds a review opinion to a model with millimeter-level precision and automatically generates a structured comment, generates a graded rectification task and pushes the task to a person in charge in real time after keyword extraction and risk library matching, synchronizes processing data to a risk library, a task system and a completion database once the data is collected, eliminates manual reporting lag, version disconnection and information silos, forms a digital progress closed loop that is auditable, traceable and person-responsible, and significantly improves progress control efficiency and data integrity.

[0028] The scheme of the application will be further described below in combination with the following specific embodiment. In the embodiment, an engineering construction project digital progress management system 20 can include five modules in Table 1, the modules work cooperatively to realize whole-process management from model uploading to risk closed loop, and the modules, composition and functions are shown in Table 1.

[0029] Table 1 The system deeply integrates BIM (Building Information Modeling) technology, WebGL visualization technology and project management process, and is mainly applied to the design review stage of large engineering construction projects such as petroleum and chemical industry, long-distance pipeline and LNG engineering. Through the innovative lightweight engine and structured data binding technology, three core functions of zero-client review, real-time collaborative annotation and risk intelligent early warning of the design model are realized.

[0030] Specifically, the digital progress management system 20 of the engineering construction project can include: A model uploading and lightweight module 210, configured to receive a three-dimensional model file of an engineering to be audited uploaded by a user, perform format analysis and patch simplification on the three-dimensional model file, and obtain a lightweight model; The three-dimensional model file of the engineering to be audited refers to a three-dimensional digital engineering model that has not completed design review, has not been officially approved, is in the review or signing stage, and is used for collision checking, specification compliance verification and multi-party collaborative review in the browser-side rendering module. For details, please refer to Figure 4 The model schematic diagram shown in the figure.

[0031] The core function of the model uploading and lightweight module 210 is to realize the cross-platform receiving and standardization processing of industrial-grade three-dimensional model files (DWG / STEP / IGES, etc.); through geometric topology analysis and patch simplification algorithm, the original model is compressed by more than 70%; the glTF2.0 format compatible with WebGL is output, which meets the browser-side rendering requirements.

[0032] Optionally, the model uploading and lightweight module is specifically configured to: Determine the original file format of the three-dimensional model file; in the scheme of the present application, more than 30 original file formats are supported.

[0033] Convert the three-dimensional model file into geometric data of a unified data structure according to the original file format; specifically, the three-dimensional model file can be converted into geometric data of a unified data structure through a converter.

[0034] Triangular patch simplification processing is performed on the geometric data to obtain a simplified triangular mesh model that retains key engineering features and reduces data volume; specifically, the QEM algorithm can be used to perform triangular patch simplification processing on the geometric data, The simplified triangular mesh model is compressed to obtain a lightweight model, which can also be referred to as a glTF file.

[0035] Optionally, when the model uploading and lightweight module converts the three-dimensional model file into geometric data of a unified data structure according to the original file format, it is specifically configured to: If the original file format is a native format, the three-dimensional model file is directly subjected to geometric topology analysis to obtain geometric data of a unified data structure; If the original file format is an intermediate format, the three-dimensional model file is standardized by a converter to obtain geometric data with a unified data structure.

[0036] The intermediate format is a general exchange format file that is not native and needs to be decoded by a converter, including but not limited to STEP, IGES, STL, and OBJ.

[0037] Optionally, the model uploading and lightweight module is used to compress the simplified triangular mesh model to obtain a lightweight model. The improved QEM algorithm is used to perform secondary error measurement on the simplified triangular mesh model to form a re-compressed grid. The QEM algorithm can be improved to perform secondary error measurement on the simplified triangular mesh model, and the simplification rate is limited to ≤10% in the welding node and valve interface area according to the feature retention factor, and the simplification rate is up to 85% in the remaining area to form a re-compressed grid. The BasisUniversal super compression algorithm is used to convert the model texture corresponding to the simplified triangular mesh model from the original format to the target format to obtain GPU direct decompression texture data. The original format refers to the bitmap format used by the map file before being processed by the system, and the target format refers to the compression format supported by the GPU natively, such as ETC1S / BC7. Non-geometric data in the simplified triangular mesh model is removed to obtain engineering metadata, which is encoded according to the JSON-LD standard to obtain structured metadata. According to the re-compressed grid, the texture data and the structured metadata, a lightweight model is generated.

[0038] Based on the above content, see Figure 3 The technical path of the model uploading and lightweight module includes: 1. Multi-format analysis engine architecture: The "layered analysis + plug-in adaptation" architecture supports the input of three-dimensional model files in 30+ industry formats (DWG / STEP / IGES / SLDPRT, etc.). The bottom layer is based on the OpenCASCADE kernel to build a geometric topology analysis framework, and dedicated analysis plugins are designed for different formats: For AutoCAD's DWG format, the internal binary structure (such as the DXF format mapping table) is restored through reverse engineering to directly extract geometric metadata such as layers, block references, and dimension annotations, avoiding information loss caused by intermediate format conversion. For STEP neutral format, the product data structure defined by its EXPRESS language is used to convert entity topology (such as faces, edges, and vertices) to a system-recognizable geometric data structure (Vertex / Face / Edge objects) through a syntax parser. For the SLDPRT format of SolidWorks, a lightweight parser encapsulated by WebAssembly (based on SwDocumentManagerSDK) extracts feature trees and parameterized constraints, preserving the editable data of the model. During the parsing process, a "format verification and fault tolerance mechanism" can also be introduced: the file integrity is verified by hash value, and the fragment repair is automatically started for damaged files (based on redundant geometric information completion); for non-standard formats (such as custom extended formats), a machine learning model (training samples containing 100,000+ industrial models) is used to predict the file structure, and the success rate of parsing is improved to 99.2%. 2. Lightweight compression algorithm cluster: Using a three-level compression strategy of "geometry simplification + texture compression + data structuring", a compression rate of more than 70% of the original model is achieved: Geometry simplification layer: the core uses an improved QEM (quadratic error metric) algorithm, which optimizes the error function for the regular geometric features of pipelines and equipment in petrochemical models (such as cylindrical surfaces and planes), and preserves key dimensions (such as pipe diameters and equipment flange thicknesses) when simplifying triangular facets. At the same time, a "feature retention factor" is introduced to limit the simplification rate to within 10% for areas with high precision requirements such as welding nodes and valve interfaces, and the simplification rate for ordinary areas can reach 85% (such as the outer wall of large storage tanks); Texture compression layer: BasisUniversal super compression algorithm is used to convert model textures (such as equipment painting and identification) from the original format (PNG / JPG) to ETC1S / BC7, which is natively supported by GPUs, with a compression ratio of 1:8. At the same time, mipmap hierarchical generation technology can ensure the clarity of textures at different scales; Data structuring layer: redundant non-geometric data (such as design software version information and temporary cache) in the model is stripped, and only metadata related to engineering management (such as material, specification, and pressure level) is retained and stored in a JSON-LD standard structure, reducing the data volume by a further 20%. Compression process accelerated by WebAssembly: the computationally intensive geometry simplification algorithm (implemented in C++) is compiled into a wasm module, which enables parallel computing on the browser side (using WebWorker), achieving a 300% speedup over pure JavaScript implementation. The compression of a 3.2GB DWG model takes less than 4 minutes.

[0039] 3. Localization adaptation and output control: To meet the information security requirements of the energy industry, the module is deeply adapted to the domestic software and hardware ecosystem: The output format glTF2.0 adds a nationalization identification field (such as “national_standard: GB / T51212-2016”), which is associated with the classification code of the domestic BIM standard; When interfacing with the Dream Database, lightweight models are transmitted through a custom binary protocol to avoid intermediate format conversion. For domestic operating systems such as Kirin OS and UOS, the memory management strategy is optimized (such as reducing 60% of memory fragmentation) to ensure smooth processing of 10GB-level raw models on 32GB memory of domestic terminals. The module also designs a “lightweight quality evaluation mechanism” to automatically generate a quality report by calculating the geometric deviation (using the Hausdorff distance algorithm) and key dimension error (such as pipe length and device spacing) between the simplified model and the original model. Only models with a deviation of ≤0.5mm are allowed to proceed to the next step, ensuring engineering precision requirements.

[0040] The browser-side rendering module 220 is used to dynamically load lightweight models with zero plugins, providing users with interactive functions for online design and review of lightweight models across terminals. The interactive functions include rotation, sectioning, measurement, and annotation positioning. The core function of the browser-side rendering module 220 is to load GB-level models with zero plugins (≤2 seconds loading speed) and provide professional interactive functions (rotation, scaling, sectioning, and measurement) for a consistent rendering experience across terminals (PC, mobile, and tablet).

[0041] The interactive functions include but are not limited to dynamic sectioning, which adjusts the sectioning plane to update the model display in real time; Avatar roaming, which uses WASD controls for first-person perspective navigation; and component visibility, which selectively displays components based on the model tree.

[0042] Optionally, the browser-side rendering module includes: The block transmission unit is used to divide the lightweight model into multiple blocks based on spatial regions and implement the download of multiple blocks through HTTPRange requests. The local cache unit is used to cache the downloaded blocks through ServiceWorker and store the model metadata corresponding to all blocks in IndexedDB for local reading during secondary loading.

[0043] The model metadata refers to a set of lightweight information describing the non-geometric features of the block, including at least the spatial bounding box, material identifier, node level, LOD level, and visibility flag. It is used to quickly reconstruct the model index and rendering state without reanalyzing geometric data during secondary loading.

[0044] Optionally, the block transmission unit is also used to: The LOD0-3 four-level grids are configured for different components in each block of the lightweight model, and the corresponding level of grid is dynamically switched for downloading according to the camera distance and device performance, wherein the accuracy corresponding to LOD0 to LOD3 decreases in turn.

[0045] Different components refer to a set of independent components in the three-dimensional engineering model, which are divided according to design specialty, functional unit or spatial region, including but not limited to process pipeline section, equipment body, civil foundation, structural frame, electrical bridge and its attached valve, flange, hanger unit.

[0046] The LOD0-3 four-level grid refers to four-level different face number versions of the same component, LOD0 level, maintaining the original high-precision grid; LOD1-2 level, halving the triangular face number step by step; LOD3 level, face number reduction of 90% for overview of long-range / mobile terminal, forming a four-level discrete grid set from fine to coarse. Such division is to reduce GPU load and network transmission according to demand without losing key review information: Specifically, LOD0 (original precision) - for close-range review, measurement, comment positioning, ensuring millimeter-level precision; LOD1-2 (medium precision) - maintain visual continuity when browsing at medium distance or quickly rotating, halve the face number, and improve the frame rate; LOD3 (low precision) - long-range, overview, mobile terminal or low-performance device, face number reduction of 90%, significantly reducing rendering pressure, ensuring smooth operation.

[0047] Through four-level discrete grids, dynamic switching can be achieved according to camera distance and device performance, realizing the optimal balance of second-level loading, 60% GPU load reduction and cross-terminal consistency.

[0048] Optionally, the above-mentioned browser-side rendering module further comprises: A dynamic sectioning unit for performing real-time sectioning of the lightweight model on the GPU side through a WebGL fragment shader; An avatar roaming unit for providing WASD key control with a first-person camera, and realizing collision reduction between the first-person camera and the model surface of the lightweight model through ray detection; A component display and measurement unit for hiding or displaying corresponding components based on the model tree Group level according to filtering conditions, and performing precision measurement using a spatial two-point distance algorithm.

[0049] Through the component display and measurement unit, corresponding components can be hidden or displayed instantly according to user-selected filtering conditions (specialty, material, construction state, etc.) by switching the visibility attribute of the model tree Group level, to realize focused review and reduce visual interference.

[0050] Optionally, the browser-side rendering module mentioned above also includes: The device capability detection subunit is used to determine the device's GPU performance during the loading phase using WebGLRenderer.capabilities, so as to automatically adjust the rendering parameters according to the device's GPU performance. The touch interaction adaptation subunit is used to generate physical view transformation parameters on the mobile terminal based on different operation gestures, and set the operation sensitivity coefficient according to the screen size, so as to uniformly map the physical view transformation parameters to the model view transformation parameters based on the operation sensitivity coefficient. The performance monitoring and dynamic adjustment subunit is used to automatically reduce the rendering quality parameters in the rendering parameters when the frame rate is lower than the preset threshold for a first set time.

[0051] The browser-side rendering module is the core of achieving "zero client-side review," and its technical implementation focuses on "efficient loading, accurate interaction, and cross-terminal compatibility," as detailed below: 1. Second-level loading engine design: A rendering pipeline based on Three.js and WebGL 2.0, employing progressive loading and on-demand scheduling, enables the initial loading time of GB-level models to ≤2 seconds, used for loading 3D model files of projects awaiting review. Chunked transfer mechanism: The lightweight glTF model is divided into 256KB blocks according to spatial regions (such as the device area and the pipe gallery area), and the chunked download is achieved using HTTPRange requests. The model blocks within the viewport are loaded first on the first screen (determined by pre-calculated spatial index), and the remaining blocks are loaded asynchronously in the background, with real-time feedback on loading progress (accuracy to 1%). LOD dynamic scheduling: For different components in the lightweight model, multiple levels of detail (LOD0-3) are set. LOD0 is the original precision (for review), and LOD3 is simplified precision (reducing the number of triangles by 90%, for overview). It dynamically switches based on camera distance (e.g., automatically switching to LOD3 if >50 meters) and device performance (LOD2 is enabled by default on mobile devices), reducing GPU rendering pressure by 60%. Optimized caching strategy: ServiceWorker is used to cache loaded model blocks, combined with IndexedDB to store model metadata (such as bounding boxes and material information). During secondary loading, data is read directly from local storage, improving loading speed to within 0.5 seconds. For frequently accessed models (e.g., viewed ≥3 times per week), a "preload task" is automatically generated to cache updated content in advance when the system is idle. 2. Implementation of professional-grade interactive toolset: To address the review requirements of petrochemical engineering projects, a high-precision interactive tool was developed. Technical details are as follows: Dynamic sectioning function: Real-time sectioning is realized through the Fragment Shader of WebGL. After the user defines the sectioning plane (X / Y / Z axis or custom plane), the system automatically calculates the plane equation (ax+by+cz+d=0) and adds a conditional judgment in the shader: if the pixel point is on one side of the plane, it is discarded, realizing the non-delayed sectioning effect. Multiple plane combination sectioning (such as sectioning along X=10 and Y=20 at the same time) is supported, and the highlight display of the sectioning edge (line width 2px, color #FF5733) is retained; Avatar roaming control: WASD key control is realized based on the PerspectiveCamera, and collision detection is realized by combining Raycaster. Specifically, when the PerspectiveCamera moves to a model surface of the lightweight model ≤0.5 meters, automatic speed reduction (speed reduction of 50%) is triggered to avoid wall penetration. The roaming speed can be adjusted (0.1-5m / s), and the "path recording" function (save key viewpoints coordinates) is supported, and the review path can be reproduced with one key; Component display and measurement: Component filtering is realized through the model tree (based on the Group hierarchy of Three.js), which supports filtering by profession (process / civil / electrical), material (carbon steel / stainless steel), and state (completed / under construction). The measurement tool uses a spatial two-point distance algorithm ( , with an accuracy of 0.1mm, and the measurement results are automatically associated with the model metadata (such as pipe length = measurement value + flange thickness), and support for labeling and saving (bound to the model coordinates). 3. Cross-terminal rendering consistency guarantee: To achieve consistent experience on PC / mobile / tablet, the browser rendering module uses an "adaptive rendering pipeline": Device capability detection: Load the device GPU performance (such as maximum texture size, whether VAO vertex array object is supported) through WebGLRenderer.capabilities, automatically adjust the rendering parameters, which can include: turn off anti-aliasing (MSAA) on low-end devices, reduce shadow precision (shadow map size from 2048px to 1024px); Touch interaction adaptation: Mobile devices use double-finger zoom (based on pinch gesture to calculate zoom factor), single-finger rotation (map gesture direction to Euler angles), and double-click centering (move the click point to the center of the viewport), with operation sensitivity adapted to screen size (7-10 inch tablet magnification factor 1.2x); Performance monitoring and dynamic adjustment: built-in FPS monitoring module (sampling frequency is 30Hz), when the frame rate is less than 24fps, automatically start the degradation strategy (such as reducing 50% of the particle effect, turn off unnecessary reflection material), ensure smooth operation. On some tablets (such as Huawei MatePadPro), it can stably render 500 million polygons of blocks.

[0052] In the scheme of the present application, the interaction processing logic shown in Figure 6 is implemented to interact with the lightweight model.

[0053] The comment management module 230 is configured to bind the review opinions input by the user on the lightweight model to the model space coordinates to generate structured comments, and track, synchronize and archive the full life cycle state of the structured comments, so as to realize closed-loop management of design problems in the three-dimensional model file. The core function of the comment management module is to establish accurate binding (error <0.1mm) between comments (review opinions) and model space coordinates; store the comments in the full life cycle data (creation→processing→closed loop); and generate traceable review reports (including 3D position snapshots). Specifically, the interface shown in Figure 5 is used to implement the comment on the lightweight model.

[0054] Specifically, the comment management module includes: The millimeter-level positioning subunit is configured to convert the two-dimensional pixel coordinates corresponding to the review opinions input by the user on the lightweight model into model space coordinates, and correct the converted three-dimensional coordinates by precomputing the vertex offset before and after the model simplification; The structured comment subunit is configured to bind the business attributes of each block, including model ID, component ID, professional type, priority, attachment ID and reference standard clause, and store them in JSON format and transmit them in Protobuf protocol; The life cycle subunit is configured to perform three checks of responsibility person role matching claim, rectification evidence submission and supervision audit on the comment state conversion request according to the conversion rules preset in each stage of the full life cycle of the to-be-audited project, and the comment state conversion request is an electronic instruction for pushing the specified comment from the current state to the next state in the full life cycle; after the three checks are passed, the specified comment is pushed from the current state to the next state in the full life cycle, and the corresponding SHA-256 chain fingerprint log is generated; The collaboration and report subunit is configured to merge multiple user concurrent comments by using operation conversion algorithm, and automatically generate vector PDF review reports with 3D position snapshots, standard clauses and electronic seals according to the professional, problem type and processing state.

[0055] The logic for converting the two-dimensional pixel coordinates corresponding to the review opinion input by the user on the lightweight model into model space coordinates and realizing binding is as follows: Screen coordinates (two-dimensional pixel coordinates corresponding to review opinion) -> normalized device coordinates (model space coordinates); generate a ray vector -> intersect the triangular facets; raycaster.setFromCamera(mousePos, camera); const pos = raycaster.intersectObjects(model)[0].point; Data structure: interface Annotation { id: "ANN-2024-001"; position: { x: 12.34, y: 56.78, z: 90.12}; / / world coordinates viewpoint: { cameraPos: [10, 20, 30], target: [5, 5, 5] / / review perspective }; status: "PENDING" | "RESOLVED"; / / life cycle }; In the present application scheme, the annotation management module is the core of realizing "real-time collaborative review", and its technical implementation is carried out around "spatial accurate binding, full life cycle tracking, cross-role collaboration", as follows: 1. Millimeter-level spatial positioning technology: Through the accurate mapping of "screen coordinates (two-dimensional pixel coordinates corresponding to review opinion) -> world coordinates (model space coordinates)", the error of the binding of the annotation and the model is realized to be less than 0.1 mm, and the technical path is as follows: Coordinate conversion link: when the user clicks the screen, the system first obtains the click position in the DOM coordinate system (clientX / clientY), that is, the two-dimensional pixel coordinates corresponding to the review opinion, converts it into normalized device coordinates (NDC, range [-1, 1]) through Three.js Vector2, and then combines the projection matrix (ProjectionMatrix) and the view matrix (ViewMatrix) of the camera to calculate the ray vector (Ray) in the world coordinate system. Through the Raycaster.intersectObjects() method, the intersection point with the lightweight model is detected, and the intersection point coordinates are corrected at the sub-pixel level (based on triangular barycentric coordinate interpolation), and finally the positioning accuracy is improved to 0.05 mm; Space index optimization: To accelerate ray detection (especially for million-level patch models), R-tree space index is used to preprocess model patches. Specifically, the grid is divided according to the spatial region (such as a 10m x 10m x 10m cubic grid), and each grid is associated with the patch ID contained. When detecting, the ray is first collided with the AABB (axis-aligned bounding box) of the grid to filter more than 90% of irrelevant patches, and the detection efficiency is increased by 10 times. Wherein, a block contains several patches and corresponding component, material and other information. Wherein, the model patch is generated by the "model uploading and lightweight module" when analyzing the geometry topology and triangulating the three-dimensional model file: first read the NURBS, BREP or parametric surface in the file, discretize it through the OpenCASCADE kernel, and then convert it into a triangular mesh. After QEM simplification, the final million-level triangular patch for rendering is obtained, and then the R-tree index is established according to the 10m x 10m x 10m spatial grid. A block contains several patches falling into the grid and their corresponding component, material, LOD and other attribute data.

[0056] Positioning deviation compensation: For fish-eye effect (edge area distortion) and geometric deviation caused by model simplification, a "deviation compensation factor" is introduced. The vertex offset before and after model simplification is pre-calculated, and the coordinates are automatically corrected during positioning to ensure that the annotation point always falls on the original design position (such as the center of the pipe weld).

[0057] 2. Structured annotation and life cycle management: Annotation data adopts a double-layer structure of "spatial information + business attributes" to support full-process tracking: Data structure design: In addition to basic fields (id / position / viewpoint / status), business fields are expanded, including: Associated objects: modelId (model ID), componentId (component ID, such as pipe ID: PIPE-1001), professional type (process / equipment / civil engineering); Content attributes: text (annotation text), priority (priority: high / medium / low), attachmentIds (attachment ID, supporting pictures / PDF), standardRef (reference standard, such as "HSE5.2"); Process data: creatorId (creator), createTime (creation time, accurate to milliseconds), lastModifierId (last modifier), statusHistory (status change record, including timestamp and operator). Data is stored in JSON format and transmitted through Protobuf protocol (compression rate 40%).

[0058] Life cycle state machine: Design the state flow mechanism of the design to be audited project in each stage of the whole life cycle "create → pending → processing → pending confirmation → closed loop → archive", and each state transition needs to meet the corresponding conversion rule; Pending → processing: responsibility person needs to claim (through system role matching, such as pipeline professional annotation automatically pushed to pipeline engineer); Processing → pending confirmation: needs to submit rectification evidence (such as modified model screenshot, calculation book); Pending confirmation → closed loop: needs to be audited by the supervisor (audit standards are associated with project quality specifications).

[0059] State change is synchronized to all associated users in real time through WebSocket (such as after creating a comment, it is pushed to related parties within 5 seconds), and change log is recorded (non-tamperable, SHA-256 hash verification is used).

[0060] 3. Collaborative annotation and review report generation: Support real-time collaboration for multiple users and generate traceable review reports: Real-time collaboration mechanism: use "operation conversion (OT)" algorithm to handle concurrent annotations. Specifically, when multiple users annotate the same position at the same time, the system automatically merges the operations (such as A adds text, B adds attachments, and the combined content is retained), avoiding conflicts. Annotation content is broadcast to all online users through WebSocket (delay < 300ms), and the user interface updates the annotation markers in real time (different users are distinguished by different colors, such as red = design, blue = supervision); Review report automation: The report generation module extracts annotation data and model snapshots, and summarizes them according to "professional classification → problem type → processing status": 3D position snapshot: capture the current view through Three.js's renderer.domElement.toDataURL(), and automatically add annotation point highlights (red circles + arrows); Data association: bind annotations with model metadata (such as component specifications, design parameters) to generate "problem details table" (including violated standard clauses, impact scope assessment); Format adaptation: Supports exporting PDF (with vector graphics), DOCX (editable), and HTML (online viewing) formats. PDF reports use electronic signatures based on the SM2 algorithm to ensure they cannot be tampered with.

[0061] The risk conversion and task management module 240 is used for keyword extraction of structured annotations and matching with a risk library, automatic generation of corresponding risk items and rectification tasks, real-time pushing of rectification tasks to responsible persons and completion of time control and closed-loop verification of rectification tasks according to a state machine; The core function of the risk conversion and task management module 240 is to automatically identify risk keywords (such as "insufficient spacing") in the annotations, intelligently match HSE risk library items and classify them (P0 / P1 / P2), and generate to-do tasks and push them to responsible persons. Different rectification tasks correspond to different time control rules: P0-level tasks: 2 hours of response; P1-level tasks: 24 hours of processing; P2-level tasks: 72 hours of closed loop.

[0062] The management mechanism of the state machine is: pending → in processing (received by the responsible person) → solved (model modified) → closed (supervisor confirmed).

[0063] Optionally, the above-mentioned risk conversion and task management module specifically includes: A risk intelligent identification subunit is configured to perform named entity recognition and relationship extraction on structured annotations using a pre-trained keyword extraction model, obtain triples, and automatically match them with an HSE risk library, and output P0-P2 classification results according to the matching results; An intelligent task generation subunit is configured to automatically generate rectification tasks containing task names, rectification requirements, model snapshots, standard clauses, and response time based on P0-P2 classification results, wherein different levels of rectification tasks correspond to different time limits; A responsible person matching subunit is configured to push rectification tasks to responsible persons; A closed-loop verification subunit is configured to perform model version comparison, OCR evidence identification, and supervisor electronic signature acceptance in sequence after the rectification task is submitted, and complete the closed-loop verification of the rectification task.

[0064] In the scheme, the risk conversion and task management module is the core that connects "problem discovery" and "closed-loop rectification", and its technical implementation focuses on "intelligent risk identification, accurate task allocation, and time control", which is as follows: 1. NLP-driven intelligent risk identification: Based on natural language processing technology, risk information is extracted from annotations and automatically matched with an HSE library: Keyword extraction model: adopts "BERT pre-training + domain fine-tuning" architecture, training data includes 100,000+ engineering annotation texts, 5,000+ HSE specification clauses (such as GB50484-2019 "Petroleum and Chemical Engineering Construction Safety Technology Standard"). The model extracts risk elements (such as "insufficient spacing", "welding defects", "material mismatch") through named entity recognition (NER), and identifies "problem-standard-impact" triplets (such as "process pipeline and fire-fighting pipeline spacing less than 150mm - violates HSE5.2 - affects welding progress") through relation extraction; Risk classification rule engine: outputs P0-P2 classification results based on matching results, specifically, determines the risk matrix based on matching results, and automatically classifies based on the risk matrix (likelihood x impact): P0 level (fatal risk): such as "insufficient wall thickness of pressure vessel, explosion risk exists" (high possibility + impact range ≥100 people); P1 level (serious risk): such as "insufficient spacing of pipeline, leading to construction stoppage" (medium possibility + impact on construction period ≥7 days); P2 level (general risk): such as "incorrect identification, need to repaint" (low possibility + cost impact <50,000 yuan). Rules support customization (projects can add industry-specific clauses, such as "third-party damage risk" for long-distance pipelines), update cycle ≤1 hour. 2. Intelligent task generation and distribution: Based on risk information, automatically create rectification tasks and accurately push to responsible persons: Task generation logic: after risk item generation, the system automatically extracts key information (risk description, associated model ID, classification result), generates task sheet according to template: Task name: "[P1] Process pipeline and fire-fighting pipeline spacing insufficient rectification"; Task content: reference annotation original text + rectification requirements (such as "adjust pipeline orientation to ensure spacing ≥300mm"); Associated data: additional model snapshot, standard clause original text, impact assessment report; Time limit: P0 level 2 hours response, P1 level 24 hours processing, P2 level 72 hours closed loop (real-time countdown display). Responsible person matching algorithm: adopts "role-specialty-load" three-dimensional matching: Role matching: construction problems found by supervisors are preferentially assigned to construction unit technicians; Specialty matching: pipeline problems are assigned to pipeline engineers (based on job tag "pipeline design"); Load balancing: through real-time query of current task quantity of responsible persons (such as less than 3 pending tasks preferentially assigned), to avoid overload. Match accuracy rate of 95%, unmatched task automatically pushed to department head (30 minutes within the reminder).

[0065] 3. Real-time synchronization and time control: Through WebSocket and timer to ensure real-time update of task status, time control: State synchronization mechanism: adopt "publish-subscribe" mode, task state change (such as "claim" "submit rectification"), immediately push to related parties (including creator, supervisor, division leader) through WebSocket, push content contains state description, operator, timestamp, to ensure information transparency; Based on time limit setting three levels of early warning: First level warning (remaining 50% time): internal mail reminder; Second level warning (remaining 20% time): mobile APP push (vibration + ring); Third level warning (over time 1 hour): automatically copy to department manager, generate overtime report (including delay impact analysis). Early warning trigger through front-end timer (setInterval) and back-end timing task (Quartz framework) double monitoring, to avoid missing report. Closed loop verification mechanism: after rectification, the system automatically verifies: Model verification: check if the associated model is updated (through version comparison, such as whether the part coordinates are adjusted); Evidence verification: review the submitted rectification photos / detection reports (through OCR identification of key parameters, such as "distance 350mm"); Supervisor confirmation: need to click "acceptance passed" in the system (including electronic signature). Verification pass rate ≥98%, not passed items are automatically returned and re-timed, to ensure the quality of rectification.

[0066] Data archiving and synchronization module 250, for synchronizing processing data to risk library, task system and completion database, to ensure closed-loop management, processing data is structured comment, risk, task, change and completion data generated in the process of engineering closed-loop design review.

[0067] Among them, the core role of data archiving and synchronization module 250 is to establish a complete data chain of comment→risk→change→completion; realize multi-system data automatic synchronization (design change / document management); guarantee the integrity of offline environment data.

[0068] Among them, the synchronization processing logic includes: Data association: comment ID binding design change order, risk item associated with completion database.

[0069] Transmission mechanism includes: online mode: RESTful API synchronization (JWT authentication); offline mode: local cache + incremental synchronization.

[0070] Security control: SM4 encryption transmission, fine-grained RBAC permission control.

[0071] Key data flow includes: Model file: DWG→glTF (data volume 3.2GB→800MB); Annotation data: spatial coordinates + viewing angle state (compression ratio 92%); Risk items: structured 〈type, level, measures〉 tuples; Task package: task unit containing responsible person / due date / associated data; Archived files: PDF report + 3D snapshot + database record; The data archiving and synchronization module is the core of ensuring the integrity of the "full-cycle data chain", and its technical implementation focuses on "multi-system collaboration, secure storage, and offline availability", as follows: 1. Full-link data association architecture: Build an association network of "annotation-risk-task-change-completion" to achieve data traceability: Association model design: adopt a hybrid mode of "primary key + tag association": Primary key association: annotation ID (ANN-XXX) is associated as a foreign key to risk item (RISK-XXX.anno_id), risk item is associated to task (TASK-XXX.risk_id), task is associated to design change order (DC-XXX.task_id), forming a rigid link; Tag association: associate cross-system data (such as welding detection report and pipeline model) through unified tags (such as "project ID: LNG-2024" "professional: process"), support multi-dimensional retrieval (such as "find all detection reports related to P1 level risk"). Data chain visualization: store association relationships through Neo4j graph database, and use D3.js to draw data chain graph (nodes are data entities, edges are association types) on the front end, users can click any node to trace the full link (such as reverse view "original annotation" from "completion data"). 2. Distributed storage and synchronization mechanism: Based on Dameng database and distributed file system, realize data safe storage and efficient synchronization: Storage architecture: structured data (annotation, task, risk item) is stored in Dameng database, adopts sharding strategy (sharding by project ID), automatically expands when single table data volume ≥100 million rows; Unstructured data (model files, reports, photos) are stored in a distributed file system (such as MinIO) and are stored in a "project / specialty / time" path hierarchy (such as "LNG-2024 / process / 202408 / model.gltf"). Supports resuming and second-level retrieval. Cache layer: Redis is used to store hot data (such as tasks in the past 7 days and frequently accessed model metadata), and the query response time is ≤100ms. Synchronization strategy: Online synchronization: Cross-system data exchange is achieved through RESTful API, and the interface uses JWT authentication (validity period of 2 hours, including user role permissions), and data transmission uses SM4 encryption (key is updated dynamically every hour). Incremental synchronization: Based on data version number (such as v1.0→v1.1), identify changes, only transfer the difference (such as only synchronize the modified parts of the model, not the full file), and the synchronization efficiency is improved by 80%. Conflict resolution: When multiple systems modify the same data at the same time, use "timestamp + priority" strategy, modify and priority high (such as supervision modification > construction modification) operation covers the former, conflict record is automatically archived (including before and after modification snapshot). 3. Localization security and offline adaptation: In response to the information security requirements of the energy industry, strengthen security mechanisms and support offline scenarios: Security control: Permission management: Based on the RBAC model, refine the granularity of permissions (such as "only view comments" and "modify task status"), support data-level permissions (such as construction units can only view their own tasks). Audit log: Records all operations (login, data access, modification), including IP address, device information, operation content, log retention ≥6 months, supports "user / time / operation type" retrieval; Third-level protection adaptation: Through log auditing, intrusion detection, and data encryption (SM4 for storage encryption and TLS1.3 for transmission encryption), it meets the third-level protection requirements and passes the localization security evaluation. Offline synchronization: Local cache: Mobile devices use SQLite to store offline data (such as pending tasks and downloaded models), and support offline adding comments (store locally, mark "to be synchronized"). In order to better illustrate and understand the principles of the method provided by the present application, the scheme of the present application is described below in combination with an optional specific embodiment. It should be noted that the specific implementation of each step in the specific embodiment should not be understood as a limitation on the scheme of the present application. On the basis of the principles of the scheme provided by the present application, other implementation manners that can be thought of by those skilled in the art should also be regarded as within the protection scope of the present application.

[0072] In a certain LNG receiving station construction project, the design unit uploads a three-dimensional model of the station field process pipeline (DWG format, original file size 3.2 GB) through the system, and the model uploading and lightweight module immediately starts the automatic processing flow. The system first identifies the file format and verifies the integrity, then calls the OpenCASCADE kernel for geometric topology analysis, performs triangle facet simplification processing through the QEM algorithm, reduces the number of facets from 1.2 billion to 36 million, and simultaneously performs texture compression processing (compression ratio 1:8), and finally outputs a glTF2.0 format file (size reduced to 800 MB). The entire process takes about 3 minutes and 28 seconds, meeting the processing requirements of large engineering models.

[0073] The project progress management module imports the WBS progress plan prepared by Primavera P6 through a standard interface, and the system automatically establishes a dynamic association between the model components and the construction tasks, such as binding the process pipeline welding task to the pipeline component ID in the model. The three-dimensional visualization module maps the progress status in real time to the model display: the pipeline welding task completed on schedule is displayed in green, and the valve installation task that has lagged behind for 15 days is displayed in red. Engineers can directly view the lag reason analysis report (including responsible unit, lag days, impact scope, etc.) by clicking on the red area in the model.

[0074] During the detailed model review phase, engineers found that the process pipeline and the fire pipeline had a design conflict of insufficient 150mm spacing at coordinates (X:12.34, Y:56.78, Z:90.12), and immediately added a structured comment using the comment management module: "Process pipeline and fire pipeline spacing insufficient 150mm, violating HSE specification article 5.2, will affect subsequent welding construction progress". The system NLP engine extracts the keyword "insufficient spacing" in real time, automatically matches the "safety spacing violation" clause in the HSE risk library, generates a P1-level risk item (code: RISK-2024-038) and pushes it to the mobile terminal of the pipeline design engineer. The system synchronously creates a to-do task, requiring the rectification to be completed within 72 hours, and notifies the relevant supervisors through the internal message.

[0075] The pipeline design engineer receives the mobile terminal alarm and immediately initiates design optimization, adjusts the pipeline route and updates the three-dimensional model. After submitting the modification in the system, the progress management module automatically updates the completion rate of the related task (from 65% to 82%), and the three-dimensional visualization module synchronously updates the collision point state in the model from red to green. The system automatically associates the comment ID (ANN-2024-001) to the design change order (DC-2024-005), and synchronously archives the processing process, new model and detection report to the completion database, forming a complete closed-loop management record. The entire processing cycle from problem discovery to closed-loop confirmation only takes 68 hours, which is 127% more efficient than the traditional process.

[0076] During the execution of a certain long-distance pipeline project, the construction party found that the design drawings conflicted with the site terrain during welding operations, and immediately initiated a design change application through the system. The system automatically executes the change management process: first, it freezes the current three-dimensional model version (V1.2) and generates a version snapshot, then it calls the weight distribution algorithm in the progress management module to evaluate the impact of the change. After calculation, the change causes the trench excavation task weight to increase from 15% to 22%, the critical path is delayed by 7 days, and it is expected to affect 3 downstream processes.

[0077] The design manager reviews the change application in the system and directly calls the three-dimensional model for visual verification, adding a comment in the model conflict area (coordinates X:45.67, Y:89.01, Z:12.34): "Suggest adjusting the pipeline elevation +1.5m to avoid rock layers", and associating the change order number DC-2024-008. The progress management module automatically adjusts the plan according to the change content: 1) recalculate the critical path; 2) update the milestone nodes; 3) push progress warnings to related construction units. The system automatically generates a change impact analysis report, quantitatively showing a 7-day progress deviation, a 230,000 yuan cost increase, and a safety risk level upgrade to P1.

[0078] The document management module synchronously starts the archiving process: 1) automatically captures the change application form (PDF format); 2) generates a model version comparison report (including three-dimensional difference highlighting); 3) associates the welding detection data record; 4) stores it in the completion database according to the predefined coding rules (project number-professional-sequence number). During the construction phase, the supervisor checks the progress execution through the pipeline three-dimensional visualization module, and by sliding the time axis control, he can compare the differences between the 30th day plan progress (blue display) and the actual progress (green display). The system automatically marks the lagging area and displays the deviation analysis data.

[0079] After the change is implemented for three months, in the project audit stage, the management personnel quickly retrieve the change order DC-2024-008 through the system trace function, and the system automatically associates and displays: 1) comparison of the original design model and the changed model; 2) progress influence analysis curve; 3) relevant welding detection report; 4) supervision and acceptance record. The finally generated structured change report is synchronized to the group knowledge base, and the 134th pipeline rock stratum avoidance solution is supplemented and improved, providing a reference basis for subsequent projects.

[0080] The system realizes the breakthrough of engineering management through three technical innovations: 1) the three-dimensional model-progress dynamic binding technology is created for the first time, a one-to-one mapping relationship between model component ID and WBS task code (mapping accuracy ±0.1mm) is established, and automatic visual early warning of construction lag is realized; 2) a closed-loop data chain of annotation-risk-progress is constructed, ensuring that 100% of design problems are traced back to the construction stage, and the historical problem reuse rate is increased by 40%; 3) a domestic safety architecture is adopted, progress data is stored in encrypted form based on the Dream database, and a lightweight engine runs in the Kylin OS environment, and passes through the third level of information security protection certification. In terms of engineering application, the system realizes the following in the Sinopec Guangdong Petrochemical Project: 1) design review efficiency is improved by 189% (52 files processed per day); 2) quality accidents are reduced by 67%; 3) handover period of completion data is shortened by 51.1% (45 days to 22 days); 4) design change rate is reduced by 67.6% (34% to 11%). Through the accumulation of 1.3 million cases in the knowledge base, the system can provide intelligent suggestions for new projects, such as automatically recommending pipeline avoidance solutions and optimizing welding parameters, etc., and continuously improve the level of digital engineering construction.

[0081] Compared with the prior art, the scheme of the present application has the following beneficial effects: The three-dimensional model online review and comment system brings breakthrough progress to the field of engineering construction: through the lightweight engine of the browser end, the industrial-level model is loaded in seconds (<2s), which is 84% more efficient than traditional CAD software, the daily processing capacity of engineers increases from 18 to 52, and the efficiency increases by 189%; the millimeter-level spatial positioning technology (error <0.1mm) makes the accurate identification rate of pipeline collision and other problems reach 100%, combined with the automatic conversion mechanism (NLP keyword matching HSE library) of comment risk, the early detection rate of design conflict is increased by 75%, which promotes the reduction of construction quality accidents by 67%. The closed loop of "comment → risk → task → archiving" constructed by the system compresses the problem processing cycle from 72 hours to 9 hours, and realizes seamless linkage of design, supervision and construction through cross-role collaboration architecture, real-time early warning push of mobile terminal makes the response speed increase by 300%, and the supervision and acceptance efficiency improves by 120%. Finally, in the field of engineering effectiveness, the design change rate is reduced by 67% (34%→11%), the rework cost is reduced by 67.4% (2.3 million→0.75 million / item), and the handover cycle of the delivery data is shortened by 51.1% (45 days→22 days).

[0082] The above description is only the preferred embodiment of the present application and the explanation of the applied technical principles. Those skilled in the art should understand that the disclosed range of the present application is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the disclosed concept. For example, the above features are replaced with the technical features disclosed in the present application (but not limited to) with similar functions to form a technical solution.

Claims

1. A digital progress management system for engineering construction projects, characterized in that, include: The model upload and lightweighting module is used to receive the 3D model file of the project to be reviewed uploaded by the user, and to perform format parsing and facet simplification on the 3D model file to obtain a lightweight model. The browser-side rendering module is used to dynamically load the lightweight model without plugins, and provides users with interactive functions for cross-terminal online design review of the lightweight model. The interactive functions include rotation, sectioning, measurement and annotation positioning. The annotation management module is used to bind the review comments entered by the user on the lightweight model with the model space coordinates to generate structured annotations, and to track, synchronize and archive the full life cycle status of the structured annotations, so as to realize closed-loop management of design issues in the 3D model file. The risk conversion and task management module is used to extract keywords from the structured annotations and match them with the risk library, automatically generate risk items of corresponding levels and rectification tasks, push the rectification tasks to the responsible persons in real time, and complete the timeliness control and closed-loop verification of the rectification tasks according to the state machine. The data archiving and synchronization module is used to synchronize the processed data to the risk database, task system and as-built data database to ensure closed-loop management. The processed data consists of structured annotations, risks, tasks, changes and as-built data generated during the closed-loop design review of the project to be reviewed.

2. The system according to claim 1, characterized in that, The model uploading and lightweighting module is specifically used for: Determine the original file format of the 3D model file; Based on the original file format, the 3D model file is converted into geometric data with a unified data structure; The geometric data is simplified by triangular facets to obtain a simplified triangular mesh model that retains key engineering features and reduces the amount of data. The simplified triangular mesh model is compressed to obtain a lightweight model.

3. The system according to claim 2, characterized in that, When the model uploading and lightweighting module converts the 3D model file into geometric data with a unified data structure according to the original file format, it is specifically used for: If the original file format is a native format, the three-dimensional model file will be directly subjected to geometric topology parsing to obtain geometric data with a unified data structure; If the original file format is an intermediate format, the 3D model file will be standardized by a converter to obtain geometric data with a unified data structure.

4. The system according to claim 2, characterized in that, When the model upload and lightweighting module compresses the simplified triangular mesh model to obtain a lightweight model, it is specifically used for: An improved QEM algorithm is used to perform a quadratic error metric on the simplified triangular mesh model to form a recompressed mesh; The BasisUniversal supercompression algorithm is used to convert the model texture corresponding to the simplified triangular mesh model from the original format to the target format, so as to obtain texture data that can be directly decompressed by the GPU. Non-geometric data is removed from the simplified triangular mesh model to obtain engineering metadata. The engineering metadata is then encoded according to the JSON-LD standard to obtain structured metadata. A lightweight model is generated based on the recompressed mesh, the texture data, and the structured metadata.

5. The system according to any one of claims 1 to 4, characterized in that, The browser-side rendering module includes: The chunked transmission unit is used to divide the lightweight model into multiple chunks according to spatial regions, and to use HTTPRange requests to download multiple chunks in chunks. The local cache unit is used to cache downloaded blocks through ServiceWorker and store the model metadata corresponding to all blocks in IndexedDB for local reading during secondary loading.

6. The system according to claim 5, characterized in that, The segmented transmission unit is further configured to: The lightweight model configures LOD0-3 four-level grids for different components in each block, and dynamically switches the corresponding grid level for downloading within the same block according to the camera distance and device performance. The accuracy of LOD0 to LOD3 decreases sequentially.

7. The system according to any one of claims 1 to 4, characterized in that, The browser-side rendering module also includes: A dynamic slicing unit is used to perform real-time slicing of the lightweight model on the GPU via a WebGL fragment shader; The Avatar roaming unit is used to provide WASD key control with a first-person camera, and combined with ray detection to achieve collision deceleration between the first-person camera and the model surface of the lightweight model. The component display and measurement unit is used to hide or show corresponding components based on the model tree Group hierarchy according to filtering conditions, and to perform accuracy measurement using a spatial two-point distance algorithm.

8. The system according to any one of claims 1 to 4, characterized in that, The browser-side rendering module also includes: The device capability detection subunit is used to determine the device's GPU performance during the loading phase using WebGLRenderer.capabilities, so as to automatically adjust the rendering parameters according to the device's GPU performance. The touch interaction adaptation subunit is used to generate physical view transformation parameters on the mobile terminal based on different operation gestures, and set the operation sensitivity coefficient according to the screen size, so as to uniformly map the physical view transformation parameters to model view transformation parameters based on the operation sensitivity coefficient. The performance monitoring and dynamic adjustment subunit is used to automatically reduce the rendering quality parameter in the rendering parameters when the frame rate is lower than a preset threshold for a first set duration.

9. The system according to claim 5, characterized in that, The annotation management module specifically includes: The millimeter-level positioning subunit is used to convert the two-dimensional pixel coordinates corresponding to the review comments input by the user on the lightweight model into model space coordinates, and correct the converted three-dimensional coordinates by pre-calculating the vertex offsets before and after model simplification. The structured annotation subunit is used to bind business attributes, including model ID, component ID, professional type, priority, attachment ID and referenced standard clauses, to each block, and store them in JSON format and transmit them using the Protobuf protocol. The lifecycle subunit is used to perform a triple verification of the annotation status transformation request in sequence, based on the preset transformation rules of the project to be reviewed at each stage of the entire lifecycle: matching and acknowledging the responsible person's role, submitting rectification evidence, and supervising review. The annotation status transformation request is an electronic instruction that requires a specified annotation to be advanced from the current state to the next state in the entire lifecycle. After all three verifications are passed, the specified annotation is advanced from the current state to the next state in the entire lifecycle, and a corresponding SHA-256 chain fingerprint log is generated. The Collaboration and Reporting subunit is used to merge concurrent comments from multiple users using an operation conversion algorithm, and automatically generate vector PDF review reports with 3D location snapshots, standard clauses, and electronic signatures according to profession, issue type, and processing status.

10. The system according to any one of claims 1 to 4, characterized in that, The risk conversion and task management module specifically includes: The risk intelligent identification subunit is used to perform named entity recognition and relation extraction on the structured annotation using a pre-trained keyword extraction model, obtain triples and automatically match them with the HSE risk database, and output P0-P2 classification results based on the matching results. The intelligent task generation subunit is used to automatically generate rectification tasks containing task names, rectification requirements, model snapshots, standard clauses and response time limits based on the P0-P2 classification results. The time limits for rectification tasks of different levels are different. The responsible person matching subunit is used to push the rectification task to the responsible person; The closed-loop verification subunit is used to sequentially perform model version comparison, OCR evidence recognition, and electronic signature acceptance by the supervisor after the rectification task is submitted, thereby completing the closed-loop verification of the rectification task.

Images