A power generation engineering three-dimensional design collaborative management and control method, system, medium and device

Abstract

The present application relates to power generation engineering technical field, provide a kind of power generation engineering three-dimensional design collaborative management and control method, system, medium and equipment, comprising: through two-way data channel and data homogenization health index real-time monitoring and actively guarantee data quality, from source suppress the information silo and rework caused by heterogeneous;And introduce power generation engineering professional rule knowledge base to carry out semantic filtering on geometric collision, significantly reduce false alarm rate;At the same time, construct multidimensional quantitative evaluation system, coupled with solving progress rate and dynamic self-adaptive threshold, form rigid logic gate, force binding model quality state and flow node, eliminate quality risk downward, realize management and control automation and objectification.

Description

Technical Field

[0001] This invention belongs to the field of power generation engineering technology, and in particular relates to a method, system, medium and equipment for collaborative management of three-dimensional design of power generation engineering. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In the field of power generation engineering, design collaboration based on Building Information Modeling (BIM) is becoming increasingly popular. However, the current technology has significant shortcomings that directly affect design efficiency and quality: First, there is insufficient data homogenization. Different disciplines use different 3D design software and data standards, making it difficult to form a unified and reliable data benchmark, resulting in information silos and rework. Second, the false alarm rate of collision detection is too high. Traditional pure geometric collision detection generates a large number of "false positive" reports that conform to engineering specifications (such as support and hanger connection points), wasting the designer's review time. Third, there is a disconnect between quality control and business processes. The integrity and standardization of the 3D model cannot be forcibly linked to collaborative workflow nodes such as its release and data submission, making it easy for quality risks to flow downstream. Fourth, process triggering relies on manual intervention and is delayed. Status indicators such as model progress and defect resolution rate cannot automatically trigger subsequent collaborative tasks, resulting in passive and inefficient management.

[0004] To address the aforementioned issues, some research and practices have been conducted on online collaboration platforms. However, their improvements mainly focus on the online approval of workflows and the cloudification of communication records. They have not deeply addressed the issue of the common source of design data flow, nor have they established an active conflict identification and closed-loop verification mechanism based on intelligent algorithms. They still fall into the category of simply transferring offline models online and cannot solve the technical problems in the collaboration of 3D design (computer-aided design CAD / building information model BIM) of power generation engineering, such as the high false alarm rate of collision detection caused by data heterogeneity, the strong reliance on manual labor in quality assessment and workflow, and the passive and unresponsive control mechanism. Summary of the Invention

[0005] To address the technical problems mentioned above, this invention provides a collaborative management method, system, medium, and equipment for three-dimensional design of power generation projects. It uses a two-way data channel and a data homogenization health index for real-time monitoring and proactive data quality assurance, curbing information silos and rework caused by heterogeneity at the source. Furthermore, it introduces a professional rule knowledge base for power generation projects to semantically filter geometric collisions, significantly reducing false alarm rates. Simultaneously, it constructs a multi-dimensional quantitative evaluation system, coupling progress rate with dynamic adaptive thresholds to form rigid logic gates, forcibly binding model quality status and flow nodes, preventing quality risks from shifting downwards, and achieving automated and objective management.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of this invention provides a collaborative management and control method for three-dimensional design of power generation projects, comprising: A two-way data channel is established with the upper-level design system or the lower-level design database to synchronize project data to the local machine. During the data synchronization process, the data homogeneity health index is monitored to determine whether the data protection protocol is triggered. The data homogeneity health index integrates data structure consistency, content synchronization timeliness, and data source availability. After completing the project data synchronization, in response to the collision check command, all triangular surface patches in the model to be checked are inspected to form an original collision list. Each collision item in the original collision list is traversed, and the corresponding judgment rules in the power generation engineering professional rule knowledge base are matched and executed. The collision items in the original collision list are filtered to obtain valid collision items and push them to complete the collision rectification. In response to the model transfer instruction, a comprehensive evaluation of the model to be transferred is performed to obtain the comprehensive credibility value of the current technical status of the model. Valid collision items and collision rectification results associated with the model to be transferred are retrieved, the resolution progress rate is calculated, and the comprehensive credibility value of the current technical status of the model and the resolution progress rate are integrated to give a decision on whether to allow or reject the transfer.

[0007] Furthermore, it also includes: comparing the pushed valid collision items with the results of manual verification to obtain the comparison results, calculating the global periodic reward function, and adjusting the parameters and rules in conjunction with the reinforcement learning algorithm.

[0008] Furthermore, the overall reliability value of the current technical state of the model is: ; in, The geometric robustness score is obtained by weighting the results after counting the number of pass, warning, and failure items through model closure checks, invalid geometry checks, and normal consistency checks. The attribute information completeness score is obtained by statistically analyzing the fill rate of required attribute fields for all components in model k; operational risk entropy. M represents the total number of preset operation types. For the first observation window The proportion of the number of occurrences of a particular type of operation out of the total number of operations; , , ,and Weighting coefficients; Version evolution normativity score =(Stability Score × 0.4 + Annotation Quality Score × 0.4 + (1 - Rollback Ratio) × 0.2); where the stability score is calculated as follows: extract the timestamp sequence of the most recent N model k submissions, and calculate the standard deviation of the interval between two adjacent submissions. Compared to average submission interval The ratio of these ratios is used to obtain the stability score through a normalized mapping. The calculation logic for the annotation quality score is as follows: a thesaurus containing four types of required keywords, namely, modification location, reason for change, associated volume, and collision number, is preset. The percentage of annotations that hit all required keywords in the most recent N submissions is counted and directly used as the annotation quality score.

[0009] Furthermore, the resolution progress rate is: ; in, The total number of collisions that have been closed-looped among all valid collision items with component associations in model k; This represents the total number of all valid collisions that have a component relationship with model k.

[0010] Furthermore, the specific steps for making a decision on whether to release or reject the application are as follows: The overall credibility value of the current technical state of the model based on each model k. Calculate the average credibility of all uploaded models in the current project. And based on average confidence level Generate dynamic decision thresholds ;in, As a static base threshold, This is the negative feedback adjustment coefficient. This is the historical average credibility benchmark for the project; Based on the overall credibility value of the model's current technological state and progress rate The comprehensive intelligent decision-making score published by the computational model ;in, and These are the weighting coefficients; The decision to approve or reject a request is determined by a rigid and unavoidable logic gate: ;in, For logical AND.

[0011] Furthermore, the data homogenization health index is: ; in, To ensure data structure consistency, a weighted average of the similarity between the local copy of each data source at time point t and the authoritative copy of the data source is calculated. To ensure timely synchronization of content; To ensure data source availability, the validity of status codes and Secure Sockets Layer certificates is combined. To penalize the sensitivity coefficient, Penalty for abnormal frequency; , and These are the weighting coefficients.

[0012] Furthermore, the calculation steps for the frequency anomaly penalty are as follows: A preset synchronization period table stores the preset synchronization interval rules for all data sources; Build a log table of actual synchronization attempts to record the execution information of each synchronization task in full; Within the statistical period, the preset synchronization period table is traversed. For each data source, the total number of synchronizations that should be completed within the statistical period is calculated. That is, the theoretical number of synchronizations for a single data source = total duration of the statistical period / preset synchronization interval of the data source. The theoretical number of synchronizations for all data sources is summed to obtain the total theoretical number of synchronizations. ; Filter the records in the actual synchronization attempt log table that fall within the current statistical period and are marked as successful, and sum them up to get the total number of successful synchronizations across all data sources. ; Calculate frequency anomaly penalty .

[0013] A second aspect of the present invention provides a three-dimensional design collaborative management and control system for power generation projects, comprising: The data homogenization and rule configuration module is configured to: establish a two-way data channel with the upper-level design system or the lower-level design database, synchronize project data to the local machine, and monitor the data homogenization health index during the project data synchronization process to determine whether the data protection protocol is triggered; the data homogenization health index integrates data structure consistency, content synchronization timeliness, and data source availability; The collaborative execution and status acquisition module is configured to: after completing the project data synchronization, respond to the collision check command, check all the triangular surface patches in the model to be checked, form an original collision list, traverse each collision item in the original collision list, match and execute the corresponding judgment rules in the power generation engineering professional rule knowledge base, filter the collision items in the original collision list, obtain the valid collision items and push them to complete the collision rectification. The dynamic association verification and control module is configured to: respond to the model transfer instruction, comprehensively evaluate the model to be transferred, obtain the comprehensive credibility value of the current technical status of the model, retrieve the valid collision items and collision rectification results associated with the model to be transferred, calculate the resolution progress rate, integrate the comprehensive credibility value of the current technical status of the model and the resolution progress rate, and give a decision to allow or reject the transfer.

[0014] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described collaborative management and control method for three-dimensional design of a power generation project.

[0015] A fourth aspect of the present invention provides a computer device including a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, wherein the processor executes the program to implement the steps of the above-described collaborative management and control method for three-dimensional design of power generation engineering.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention establishes a two-way data channel and pioneers a data homogenization health index (integrating structural consistency, timeliness, and availability), enabling real-time quantitative monitoring and proactive protection of project data quality. This changes the traditional platform's "only transmits, doesn't care" approach, curbing information silos and rework issues caused by data heterogeneity at the source, and ensuring the reliability and authority of various professional design benchmarks.

[0017] This invention introduces a knowledge base of professional rules for power generation engineering to perform semantic-level intelligent filtering on pure geometric collision results. This invention effectively eliminates a large number of "false positive" reports that conform to engineering specifications (such as support and hanger connection points). Compared with traditional pure geometric detection, it can accurately identify "valid collision items", significantly reducing the amount of invalid review work for designers and making collision rectification more targeted and meaningful for engineering practice.

[0018] This invention proposes a comprehensive credibility value for the current technical state of a model. This value integrates geometric robustness, attribute completeness, version evolution standardization (including submission stability and annotation quality), and operational risk entropy, achieving an objective digital characterization of model quality. By combining the resolution progress rate (collision closure ratio) with a dynamic adaptive decision threshold, a rigid logic gate is formed, which forcibly links the quality status with the flow nodes such as release and resource submission, preventing the spread of quality risks to upstream and downstream, and realizing the automation and objectification of control.

[0019] This invention compares the pushed valid collision items with the results of manual verification, introduces a global periodic reward function and reinforcement learning algorithm, and can continuously optimize internal parameters based on historical review feedback. It has dynamic adaptability and no longer relies on lagging manual judgment. This allows the collision filtering rules and evaluation models to evolve with engineering practice, greatly improving the efficiency and intelligence level of collaborative management.

[0020] This invention introduces the standard deviation of the commit interval (to measure the stability of the development rhythm) and the hit rate of required keywords (to measure the entropy of comment information) to address the standardization of version evolution, and introduces an anomaly penalty mechanism (based on the ratio of theoretical frequency to actual success frequency) to address the frequency of data synchronization. These statistically based indicators eliminate the subjectivity and arbitrariness in the evaluation process, provide management with a fair and quantifiable basis for decision-making, and significantly enhance the traceability and controllability of the entire process. Attached Figure Description

[0021] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0022] Figure 1 This is a flowchart of a collaborative management and control method for three-dimensional design of power generation projects according to Embodiment 1 of the present invention; Figure 2 This is a flowchart of the dynamic correlation verification and rigid release decision in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the model credibility and circulation score in Embodiment 1 of the present invention; Figure 4 This is a bar chart comparing the performance of different methods in Embodiment 1 of the present invention; Figure 5 This is a schematic diagram of the structure of a computer device according to Embodiment 4 of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0024] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0025] It should be noted that the data and information obtained in this invention are all obtained with the consent of the parties involved and are legally obtained in accordance with the law.

[0026] Example 1 This embodiment provides a collaborative management and control method for three-dimensional design of power generation projects.

[0027] This embodiment provides a collaborative management and control method for 3D design of power generation projects. The core of this method lies in constructing a fully intelligent management and control system based on data homogeneity, intelligent filtering, dynamic correlation verification, and self-learning optimization. This system is deployed as software on a server cluster and provides access services to various clients via network protocols. In particular, it seamlessly integrates with mainstream industrial 3D CAD / BIM design platforms through a deeply developed application programming interface (API) and customized adapters. This not only supports the extraction of model files but also enables the acquisition of detailed model geometric data, attribute information, and operation log streams.

[0028] To achieve effective integration with CAD (Computer-Aided Design) platforms, the software adapter encapsulates dedicated data exchange protocols for different CAD software. For example, for plant 3D design software (e.g., Autodesk's AutoCAD Plant 3D or similar products), the adapter accesses the Plant3D project database and uses SQL (Structured Query Language) to extract geometric attributes of components such as pipes, equipment, and structures, including coordinates, dimensions, materials, and user-defined attributes, and serializes them into a platform-defined format. For Building Information Modeling (BIM) software (e.g., Autodesk's Revit or similar products), it extracts the geometric information and parameters of element objects through its external command mechanism and application programming interface (API), and converts them through IFC (Industrial Foundation Class, a BIM data exchange standard) intermediate format or direct memory mapping. For common anomalies during the integration process, such as API version mismatch, model file encryption, and network interruptions, the adapter has retry mechanisms, degradation schemes, and detailed error logs reported to the platform monitoring center.

[0029] This embodiment provides a collaborative management and control method for three-dimensional design of power generation projects, such as... Figure 1 As shown, the specific steps include the following: Step S100: Authoritative data homogenization configuration and dynamic quality monitoring.

[0030] In this embodiment, establishing a unique and verifiable authoritative data source throughout the entire project lifecycle is the underlying foundation for achieving efficient and accurate collaboration. Furthermore, by continuously monitoring the health of the data source through quantitative indicators, downstream risks caused by data chaos can be prevented at an early stage.

[0031] First, by connecting with upstream surveying and design project management systems (e.g., enterprise-level project planning and scheduling tools) and core CAD design data sources via API, core information such as project WBS (Work Breakdown Structure), volume structure, personnel organization, and the version and coordinate system of the initial 3D design model is synchronized, establishing a unique and authoritative data source based on industrial design data standards. Then, data consistency is quantified by calculating a data homogenization consistency index. When the index consistently falls below a safe threshold, downstream CAD model publishing and data export operations, including critical data extraction, are automatically suspended, and an alert is issued.

[0032] S101: By establishing a two-way data channel with the enterprise-level upstream design system and the underlying CAD design database, project data, including the project's work breakdown structure, volume catalog tree structure, professional design team organizational structure, personnel permission and role matrix, and key metadata of the initial design model, such as the origin coordinate system, unit system, and version number, are synchronized in real time to the local distributed cache and persistent database in two modes: fixed frequency and event triggering. This forms a dynamically updated single source of fact.

[0033] In this embodiment, specifically regarding the interface with the aforementioned enterprise-level project planning and schedule management tool, the WBS structure tree is obtained by calling its RESTful API (Representative State Transfer Application Programming Interface) according to its data model. After flattening it, it is mapped into a tree node table containing fields such as node ID, name, parent node ID, planned start time, and completion time.

[0034] For CAD model metadata synchronization, when a model version update is detected in the upstream design system, the unique identifier, coordinate system, unit system, and associated nodes of the model are obtained by listening to its release log or polling its version repository, and recorded in the relevant collision table.

[0035] The data synchronization process employs a transaction mechanism to ensure that the association between nodes and model metadata remains consistent. If synchronization fails midway, all changes will be rolled back and retried in the next cycle.

[0036] S102: To objectively assess the timeliness and accuracy of data synchronization, a data homogenization health index is designed and calculated in real time. This index is a comprehensive measure that integrates three dimensions: data structure consistency, content synchronization timeliness, and availability. Its calculation formula is as follows: ; in, For the content consistency item, its value is obtained by calculating the weighted average of the similarity between the local copy of each key data at time point t and the authoritative copy from the upstream design system and other data sources; for example, calculating the similarity of the structure and node attributes of the volume catalog tree, the mapping accuracy of the personnel permission matrix, etc. The value range is [0,1], where 1 represents complete agreement.

[0037] It should be noted that, for the volume directory tree, the tree structure and the attributes of each node are serialized into an ordered set of strings, and the Jaccard similarity coefficients of the local and upstream sets are calculated respectively; for the personnel permission matrix, the assignment status of each user-role-project triplet is compared in the local and upstream sets; and after each synchronization, a consistency verification task is started, and its execution log and intermediate calculation results are stored in the audit log table for later traceability.

[0038] in, The data source availability category combines status codes and the validity of SSL (Secure Sockets Layer) certificates. This information is recorded in a business table, and historical trends can be viewed through the management interface. Specifically, in order to obtain... The required real-time status data is periodically (e.g., every 5 minutes) sent as probe requests to the health check endpoints of each upstream system, and the returned HTTP (Hypertext Transfer Protocol) status codes, response latency, and other parameters are recorded. Certificate status, as Inputs to the calculation. Data source availability sub-item. The specific calculations are as follows: ; in, The number of requests that returned normal status codes (e.g., 2xx, 3xx) from a certain data source within the most recent statistical period; This represents the total number of requests sent to a specific data source within the most recent statistical period. This is a flag indicating the validity of the SSL certificate; 1 indicates a valid certificate, and 0 indicates an unvalid one. , Let be the weighting coefficient, satisfying + =1, typical values ​​are as follows =0.7, =0.3, at the same time, The value range is [0,1], and the closer it is to 1, the better the availability of the data source.

[0039] in, This is a frequency anomaly penalty item, its value being the deviation between the actual number of successful synchronizations and the preset frequency plan within the most recent evaluation window; if the synchronization frequency is too low or failures occur, As the value increases, the exponential decay term increases. Synchronization anomalies are quantified as penalties for health status. The sensitivity factor is penalized, with an empirical value set at 0.5.

[0040] Among them, the frequency abnormality penalty item The specific calculation steps are as follows: (1) Preset synchronization cycle table, by storing the preset synchronization interval rules of all connected data sources, for example, the project management data source is configured to synchronize once every hour, and the CAD model metadata data source is configured to synchronize once every 15 minutes; (2) Construct an actual synchronization attempt log table to record all execution information of each synchronization task, including data source identifier, synchronization attempt timestamp, and synchronization success / failure status identifier; (3) The service status and response latency of each data source API are evaluated through a heartbeat detection mechanism. The specific calculation rule is as follows: the most recent 24 hours are taken as the statistical period for synchronization. The preset synchronization period table is traversed, and the total number of synchronizations that should be completed for each data source within the statistical period is calculated. That is, the theoretical number of synchronizations for a single data source = the total duration of the statistical period / the preset synchronization interval of the data source. The theoretical number of synchronizations for all data sources is summed to obtain the total theoretical number of synchronizations. ; (4) Filter the records in the actual synchronization attempt log table that fall within the current evaluation window (i.e., the statistical period) and are marked as successful in synchronization status, and sum them up to obtain the total number of actual successful synchronizations for all data sources. ; (5) The degree of anomaly is quantified based on the deviation between the theoretical total number of synchronizations and the actual number of successful synchronizations. The formula is as follows: The value range of this sub-item is strictly locked in the interval [0,1]. The higher the value, the more serious the deviation of the synchronization frequency from the preset plan, and the more prominent the synchronization anomaly. The calculated value will be used to determine the specific values ​​of this sub-item. Substitution This allows us to quantify the impact of this synchronization anomaly on the data homogenization health index. The final penalty amount, this item's value range is strictly locked in the [0,1] interval, the closer the value is to 1, the better the availability and response performance of the data source service; for scenarios of continuous failures, It will show exponential growth, rapidly pulling down .

[0041] in, , and The weight coefficients of the above three items are respectively, satisfying... Based on engineering practice, it is set to ω c =0.6、ω f =0.2、ω a =0.2, emphasizing the accuracy of the data content.

[0042] S103: Set health alarm threshold It is usually 0.85, when For two consecutive monitoring periods, if the value is below 2 hours... At that time, the data protection protocol will be automatically triggered: First, a global alert is displayed in a prominent position on the user interface to warn that the current data benchmark may be at risk; Then, immediately freeze all high-risk operations such as the release of design models, data submission, and external export of model data that depends on the platform; Finally, a root cause analysis report is pushed to the administrator, indicating the specific abnormal data source type and the downstream business that may be affected, and mandating manual intervention for verification and repair, thereby forming a reliable self-protection logic loop at the data source.

[0043] The freeze operation is implemented at the database level by setting a flag in the relevant business table that is controlled by the global health status. When the flag is True, all related INSERT or UPDATE operations will be intercepted by the application layer and return error information.

[0044] The root cause analysis report is automatically generated by the analysis service. It queries multiple related tables and uses a decision tree algorithm to quickly locate the most likely cause. Factors contributing to the decline, such as the specific error code of the most recent synchronization failure, are analyzed, and the results are pushed to the administrator via internal enterprise communication tools.

[0045] Step S200: Design collaborative execution, multi-dimensional state acquisition and intelligent collision screening based on engineering semantics.

[0046] In this embodiment, when designers perform core collaborative operations, process data is comprehensively collected, and relevant knowledge in the engineering field is used to perform an initial intelligent screening of the massive amount of raw and noisy collision detection results, filtering out compliant collisions that are in line with the specifications but have no meaning, allowing designers to focus on real design conflicts.

[0047] In this embodiment, after completing their design work in their local or cloud-based CAD design environment through a unified portal, designers from various disciplines submit the 3D model to the platform or operate directly in the lightweight CAD view integrated into the platform and initiate a collision check command. Upon receiving the collision check command, the intelligent collision recognition engine immediately intervenes and filters the collision using a pre-built rule knowledge base trained based on the CAD design specifications of various disciplines in power generation engineering.

[0048] S201: Designers from various disciplines, such as mechanical, electrical, thermal control, and civil engineering, complete model design work in their local or cloud-based native CAD design environments. Through integrated plugins or by calling the platform's WebAPI (Web Application Programming Interface), they can submit the latest version of their 3D model to the platform's versioned model repository in a transactional manner. Simultaneously, designers can directly initiate clash checks within specified spatial ranges, such as by floor, fire compartment, or specific discipline combination, in the platform's integrated high-performance, lightweight 3D view on the web interface, according to task requirements.

[0049] S202: After receiving a collision check command, the intelligent collision recognition engine initiates a proprietary two-stage processing flow.

[0050] (1) Phase 1.

[0051] The intelligent collision recognition engine calls upon the precise geometric calculation library of the geometric calculation kernel to quickly check all triangular surface patches in the model to be inspected, accurately calculate and output all components that have spatial interference, and calculate the interference amount and collision point coordinates to form an original collision list.

[0052] Specifically, the geometric calculation kernel used is based on secondary development of open source libraries or encapsulation of commercial engines, and has been specially optimized for fast interference detection algorithms between common tubular components such as pipes, air ducts, and cable trays and solid components in BIM models.

[0053] The original collision list is a structured list containing information such as collision pair ID, globally unique identifiers of components A / B, minimum penetration depth, collision point set, and collision bounding box, and is stored in a database table.

[0054] The traditional workflow ends here, and the original collision list is directly output as a collision report.

[0055] (2) Phase Two.

[0056] Based on pre-screening using engineering rules, the intelligent collision recognition engine does not directly output the original collision list. Instead, it accesses a pre-built, dynamically loadable knowledge base of power generation engineering rules, which is coded by domain experts and contains hundreds of quantitative judgment logics. The intelligent collision recognition engine traverses each collision item in the original collision list and matches and executes the corresponding judgment rules based on information such as the type, attributes, spatial location, and system to which the collision component belongs.

[0057] In this embodiment, for the common scenario where the connection point of a pipe support overlaps geometrically with the solid support structure, since the two are geometrically bound to overlap and meet design requirements, it is considered a false positive collision. Therefore, the corresponding judgment rule in the power generation engineering professional rule knowledge base is the compliance judgment of the support point-support structure, and its quantitative judgment formula and logic are as follows: ; in, The three-dimensional coordinates of the pipe support points. The three-dimensional coordinates of the effective installation points of the solid support. Pipe support points With solid support The Euclidean distance between valid installation points; AND is the logical AND; To allow attachment determination function, return a Boolean value to determine the relationship between two components (pipe support points). With solid support Whether rigid connections are allowed depends on the system diagram, PID (piping and instrumentation diagram), or professional design.

[0058] in, and Instead of taking arbitrary points directly from the model mesh, these points are obtained by parsing the parametric information of the components or identifying specific geometric feature points. For pipe support points, the connection point attributes or specific positions on the pipe axis are obtained first. For supports, the center of the top surface or the theoretical contact point with the support surface is obtained.

[0059] The calculation of the above formula is completed in the platform's 3D geometry service, using the transformation matrix from the component's local coordinate system to the world coordinate system.

[0060] in, It is not a fixed value, but rather based on Pipe specifications, such as pipe diameter, wall thickness, medium pressure, and The type of stent is determined by querying the company's internal dynamic parameter database for the minimum, standard, or maximum permissible spacing sequence; the minimum value is used here. As a benchmark.

[0061] in, The allowable modeling and installation tolerance thresholds are defined by the project, such as ε=2mm for precision layout areas and ε=5mm for general areas.

[0062] Furthermore, maintain a standardized parameter library, stored in the form of key-value pairs or small relational tables. For example, the key is "pipe-support-minimum spacing", and the value is a specific value retrieved based on pipe diameter segmentation and pressure level. The query process is implemented within the rule execution engine.

[0063] Furthermore, when a collision item meets the conditions in the specification parameter library, the rule execution engine automatically filters it from the original collision list and records it as exempted-design compliant. This item and its judgment criteria will be stored in the audit log, will not appear in the pending task list, and will not be included in any subsequent quality assessment statistics. This process can significantly reduce the initial order of magnitude of collision reports, leaving only valid design conflicts, thus improving the operational efficiency of staff.

[0064] Specifically, the knowledge base of professional rules for power generation engineering is stored in JSON (JavaScript Object Notation) or XML (Extensible Markup Language) format files. Each rule contains fields such as ID, name, applicable component type combination, judgment logic script, and confidence weight. The rule engine loads, interprets, and executes these rules. The audit log records in detail the ID of each collision item, the ID of the applied rule, the judgment result, the calculation parameters, and the final processing status.

[0065] S203: For the remaining valid collision items after screening, instead of relying on manual allocation, the system automatically combines the model's professional attributes, volume ownership, spatial location, and design responsibility matrix to accurately push the collision to the sole or primary responsible designer. When pushing, the system will directly highlight and locate the exact collision position in the platform's 3D view and provide a one-click link to the source model file in the native CAD software, so that the designer can review or complete the closed-loop resolution.

[0066] It should be noted that highlighting and positioning are achieved by passing the collision point coordinates and components to the API of the front-end 3D rendering engine. For desktop CAD software, this is done by registering a custom URL (Uniform Resource Locator) protocol, which invokes the local CAD application and opens the specified model to position the specified primitive when the user clicks. For the lightweight view on the web, the viewpoint is switched and components are highlighted directly.

[0067] As shown in Table 1, the collision details of the circulating water pump house model show that the original number of collisions is 3 (i.e., the 3 numbers CL-4-1, CL-4-2, and CL-4-3), the number of rule filters is 1 (i.e., the result of CL-4-1 is rule filtered), the number of effective collisions is 2 (i.e., CL-4-2 and CL-4-3), and the number of closed-loop solutions is 2 (i.e., CL-4-2 and CL-4-3).

[0068] Table 1. Model Collision Details

[0069] Step S300: Integrate dynamic correlation verification of multi-dimensional state assessment with rigid release decision.

[0070] Specifically, at critical junctures where the model transitions from individual design to team collaboration or release, a mandatory, multi-factor integrated, adaptive, and unavoidable quality review is implemented. This review ensures model release quality from both technical and collaborative perspectives, while dynamically controlling project risks. Figure 2 As shown.

[0071] Specifically, when designers complete the CAD design of the model for their assigned area and submit key process requests such as model release or design information submission, the dynamic association verification and control unit will be forcibly activated.

[0072] S301: When a designer submits a model for publication or cross-disciplinary design information for approval, the dynamic correlation verification and control unit is forcibly activated. First, it calculates the comprehensive credibility value of the current technical status of the model k to be transferred. This indicator abandons the traditional binary judgment of pass or fail, and is a comprehensive evaluation based on the fusion and quantification of multi-dimensional time series data: ; in, To obtain the geometric robustness score, the API of the 3D model inspection tool is called to execute a preset set of inspection rules, including model closure checks, invalid geometry checks, and normal consistency checks. Based on the number of pass, warning, and failure items in the inspection report, a weighted score is calculated according to preset weights (e.g., pass items have a weight of 1, warning items have a weight of -0.2, and failure items have a weight of -0.5), and normalized to the [0,1] interval.

[0073] As one implementation method, To determine the version evolution standardization score, we query model k in the most recent N version commit records. The complete calculation formula for this sub-item is as follows: =(Stability score × 0.4 + Annotation quality score × 0.4 + (1 - Rollback ratio) × 0.2).

[0074] The rollback ratio refers to the ratio of the number of submissions that were withdrawn or rolled back to the previous version due to design errors, failure to pass review, or other reasons in the most recent N submissions to the total number of submissions N. The higher the rollback ratio, the more unstable the version management and the worse the standardization. Therefore, (1-rollback ratio) is used as a deduction item.

[0075] The stability score is calculated by extracting the timestamp sequence of the N most recent model submissions. Calculate the standard deviation σ between two consecutive submissions. T Compared to average submission interval The ratio of these ratios is used to obtain the stability score through a normalized mapping. The value range is locked in the [0,1] interval. A higher score indicates a more stable model submission rhythm and no abnormal situations such as sudden concentrated modifications or long-term stagnation. The calculation logic for the annotation quality score is as follows: a pre-set thesaurus containing four types of required keywords: modification location, reason for change, associated volume, and collision number; the percentage of submissions that hit all required keywords in the most recent N submissions is counted and directly used as the annotation quality score S. comment The value range is [0,1]. The higher the score, the more complete and valuable the version annotation information is.

[0076] As another implementation method, the most recent N commit records of model k can be extracted from the version management database. In a simplified implementation scenario of lightweight deployment, a minimal alternative implementation method can be adopted. = (Number of submissions conforming to the specifications) / N.

[0077] in, To score the completeness of attribute information, the fill rate of required attribute fields for all key components (such as pumps, valves, and pipes) in model k is calculated by comparing them with the preset model component attribute template. This is the fill rate. For key attributes such as KKS (Power Plant Identification System) coding, a simple validity verification will also be performed.

[0078] in, Operational risk entropy, a negative indicator, is calculated by analyzing a user's recent sequence of operations on model k (such as deleting components, redrawing pipelines, modifying system properties, etc.) to determine the variability and uncertainty of these operations, and the entropy value. The higher the value, the more chaotic the recent operations, resulting in a lower overall reliability of the model, which in turn lowers the real-time state reliability assessment value of the model. Where M is the total number of preset operation types. For the first observation window The proportion of the number of occurrences of a certain type of operation to the total number of operations, that is, the frequency of occurrence of various operation types in the observation window. Frequent deletion and undo operations will significantly increase the entropy value. This calculation is performed periodically by the platform's analysis service and the results are cached. All operation logs are collected from the CAD plugin or the platform operation interface's embedded points, providing full raw data input for the calculation of this entropy value.

[0079] in, , , and The weighting coefficients for each item are set by the project quality manager, ensuring that they add up to 1. Typical values ​​are as follows: .

[0080] It should be noted that the specific implementation involves calling the API of the model inspection service or an open-source library to run a predefined set of geometric inspection rules; checking for non-closed faces, self-intersecting faces, extremely small gaps, or free-floating points and lines in the model; the results of each inspection (pass / warning / failure) are weighted and summarized as follows: The inspection report is saved as an attachment in the model's version history.

[0081] Furthermore, It is a continuously quantified evaluation metric (range 0.0~1.0). A higher value indicates that the model is technically more reliable and ready for release. The platform continuously records the performance of each model in its database. Historical curves are used to analyze trends.

[0082] S302: Calculate the overall resolution progress rate of all valid collisions associated with model k. The valid collisions here refer to the collisions that remain after the screening in step S202 and need to be actually processed. The calculation formula is as follows: ; in, This refers to the total number of valid collisions that have been closed-loop handled after S202 intelligent filtering, where there are component associations in the model. The criteria for determining closure are: after the corresponding designer completes the collision rectification and submits supporting materials, which are verified and marked as "resolved", or after submitting a compliance exemption explanation and being reviewed and approved by the professional person in charge, all valid collisions that have been closed-loop handled are marked as "closed-loop ignored".

[0083] in, This refers to the total number of all valid collisions that are associated with the model, i.e., the total number of all real design conflicts that require manual intervention and are transferred to the processing queue after the virtual collision filtering is completed in step S202.

[0084] when When this occurs, it indicates that all design conflicts related to this model have been closed-loop resolved.

[0085] The determination of associated collisions is based on matching the IDs of component A and component B in the collision record table with the list of component IDs contained in model k. The closed-loop status of the collision is marked as resolved or ignored by the responsible designer after processing, and the reason is explained. The designer is also required to upload the modified model version or explanatory document. The life cycle of the collision is managed through a state machine.

[0086] S303: Comprehensive credibility value of the current technical state of the model based on each model k. Calculate the average credibility of all uploaded models in the current project in real time. Based on this, a dynamic decision threshold is generated. : ; in, The static baseline thresholds preset for project quality managers, such as for a nuclear power plant project with extremely high quality requirements. It can be set to 0.90; This serves as the historical average credibility benchmark for a project during its healthy and stable operation phase, and is typically obtained from statistics at the beginning of the project. The negative feedback adjustment coefficient ( The engineering significance of introducing this coefficient lies in: when the overall design of the project is healthy and the designer's output is stable. Higher than the benchmark confidence level At that time, decision threshold The quality will be appropriately reduced; however, when the overall project is in poor condition, or when rushing to meet deadlines or frequent changes may lead to a general decline in model quality. Below threshold It will automatically increase and tighten the release standards; the coefficient γ controls the sensitivity of this adjustment, and its value is usually 0.05~0.15, which needs to be determined through preliminary tests based on the project scale and risk tolerance; the value of γ can be a fixed value or dynamically adjusted according to the project stage. Its implementation is displayed to the administrator as an adjustable parameter in the platform's configuration management module.

[0087] The dynamic decision threshold is calculated by a scheduled task that queries the relevant business tables and updates them to the memory cache in order to improve decision-making efficiency.

[0088] S304: Combine the evaluation results of S301 and S302 to calculate the comprehensive intelligent decision-making score (i.e., intelligent transfer score) for this transfer application. : ; in, and Let be the weighting coefficient, satisfying The weighting settings reflect the project's value orientation. In fields like power generation engineering design, where safety and collaborative quality requirements are extremely high, a problem-solving-first strategy is typically adopted, for example, setting... or This gives the collision resolution rate A considerable or even higher weight; because even if the model itself is in good condition, as long as there are unresolved collisions, it may mean that it will have unacceptable conflicts with other disciplines, which will bring greater risks after the transfer.

[0089] S305: The final decision on whether to approve or reject the application is determined by the following rigid and unavoidable logic gates: ; That is, two conditions must be met simultaneously: Comprehensive intelligent decision-making score of the model No lower than the current dynamic decision-making threshold ; All valid collisions associated with this model must have been 100% resolved, i.e. .

[0090] If any one condition is not met, the application will be automatically and forcibly rejected, and the specific reasons will be clearly stated in the rejection notice, such as: "The overall score IFD=0.85, which does not meet the current dynamic threshold ζ=0.87" or "There are 3 unresolved collisions, which cannot be satisfied." The system provides "condition 1" and offers detailed modification guidelines and problem-solving links, thus reducing the risk of releasing low-quality models due to personal negligence or compromise.

[0091] In this embodiment, the decision logic is encoded in the gateway node of the approval workflow engine, or as an independent microservice. When a rejection is made, a notification email is automatically generated with a templated content but containing specific variable values. At the same time, in the user's task list, the rejected application will be accompanied by a link to view the reason and improvement suggestions. Clicking the link will take you to a page containing detailed scoring details and a list of unresolved collisions.

[0092] like Figure 3 As shown, this is a schematic diagram of the credibility and transfer score of the main plant thermal control instrumentation cable tray model. The horizontal axis is the model submission / evaluation time axis, and the vertical axis is the score. It can be seen that the final transfer score of the model is 0.634, and the dynamic decision threshold is 0.893. The model will be rejected.

[0093] Step S400: Global state-driven collaborative process automation and multi-dimensional visualization.

[0094] In this embodiment, based on the massive real-time data collected in step S200, the health of the project is monitored from a macro perspective, and downstream collaborative processes are automatically triggered according to preset business logic, realizing the transformation from passive response to proactive management.

[0095] Among them, the system continuously monitors macro indicators such as the completion rate of the CAD model output and the effective collision resolution rate of the entire project; when the indicators reach the threshold of the preset collaborative trigger rule base in S100, subsequent collaborative tasks are automatically created and assigned, realizing a state-driven process; all states are visualized through an integrated 3D dashboard, and managers can view the specific location of any collision point in the original CAD model.

[0096] S401: The global visualization management module aggregates and displays macro-level project indicators in real time in the form of a dashboard. These indicators are no longer simple percentages, but rather complex aggregated information systems that integrate spatial, professional, and temporal dimensions. For example, on the 3D factory model, unresolved collision density areas, areas with abnormal model update frequency, and design lag areas are highlighted with different colors. Specifically, the 3D space of the factory is meshed, and the number of collisions, model version changes, and deviations between planned and actual progress are counted within each mesh, then rendered using color gradients.

[0097] This feature relies on the thermal layer plugin for the 3D engine.

[0098] It provides a multi-dimensional comparison of the design status and collaborative quality of various disciplines.

[0099] Specifically, the data comes from aggregated queries of relevant data tables. The front end draws a radar chart, with each dimension axis representing an indicator. Lines connecting indicator values ​​from different disciplines form polygons, facilitating intuitive information comparison.

[0100] Based on historical data, short-term trend predictions are made on the number of design conflicts on the critical path and the pace of model releases, thereby assisting in management decisions.

[0101] Specifically, a simple time-series forecasting model is used to train and forecast historical indicator data. The forecast results are displayed in the form of a line chart, and confidence intervals are given.

[0102] The time-series prediction model runs periodically as a background service.

[0103] S402: The intelligent process trigger continuously monitors these macro indicators and compares them in real time with the collaborative process rule base configured in step S100.

[0104] The collaborative process rule base is pre-configured by the project manager or professional leader according to the milestone plan, and is in the form of trigger condition-execution action.

[0105] Specifically, the triggering condition can be a single indicator threshold, such as "HVAC - main plant area model progress > 80%", or it can be a compound logical condition, such as "electrical cable laying completion rate > 70%" and "civil engineering structural model stability > 0.9".

[0106] The triggering conditions are defined using expression language and stored in the core database tables.

[0107] The rules engine periodically scans these rules, queries the current project status data, and evaluates them through the aggregated view in step S401.

[0108] It needs to be explained in detail that the system automatically creates and executes actions, assigns responsible parties, and distributes task work orders for the next stage. For example, when the electrical engineering team meets the model progress target for a certain area, a task titled "Detailed Clash Check of Cable Tray and Duct System in a Certain Area" is automatically created and assigned to the HVAC team's responsible party. Specifically, task creation is achieved by calling the workflow engine's API, instantiating a predefined detailed clash check process template, and passing in parameters. The task assignment logic is based on the design responsibility matrix mentioned in S203.

[0109] Specifically, once all professional models for a fire compartment have been published and all collisions have been resolved, a notification is automatically generated that the model can be submitted to the construction drawing team, and relevant reviewers are also notified. This is achieved by checking the publication status of all models under that fire compartment and the resolution status of associated collisions. The notification is sent via an internal messaging system or email gateway.

[0110] This event-driven or state-driven process automation transforms waiting for inspection instructions into intelligent system execution, significantly shortening the waiting time in the collaborative chain and making cross-disciplinary collaboration more efficient.

[0111] Step S500: System knowledge accumulation and adaptive optimization based on reinforcement learning.

[0112] It should be specifically noted that, in order to have the ability to continuously learn and self-evolve, the platform continuously optimizes the intelligent filtering rules in S200 and the parameter weights in S300 by analyzing the differences between machine judgment and manual confirmation in historical operations, so that the platform becomes smarter with use and gradually evolves into a unique design knowledge asset for the enterprise.

[0113] In this embodiment, adaptive optimization is used not only to optimize the rule confidence of the intelligent collision screening knowledge base in S200, but also to adjust the key parameters involved in the dynamic association verification and control module in S300, including but not limited to: dynamic decision threshold. negative feedback adjustment coefficient Comprehensive intelligent decision scoring Weighting coefficients The combined credibility of β and the current technical state of the model. Weighting coefficients of each item , , and .

[0114] To achieve joint optimization of the above-mentioned multiple parameters, the optimization engine constructs reward signals for different parameter categories: For collision screening rules, the reward signal mainly comes from the comparison results between machine judgment and final human confirmation, as shown in Table 2 below and the global periodic reward function. Regarding the collision handling section; for the flow decision parameters, the reward signal comes from the actual quality feedback after the model is released, such as the rework rate, number of design changes, and collision miss rate discovered in downstream professional or construction stages. This part of the reward is in the form of performance reward items. Incorporating the form into the global periodic reward function In this process, the effectiveness of the current threshold strategy is evaluated, and relevant parameters are updated accordingly.

[0115] Through the above methods, adaptive optimization can simultaneously optimize both collision screening and flow decision points, enabling the overall decision logic of the platform to evolve collaboratively, rather than being limited to the collision filtering stage.

[0116] Among them, adaptive optimization periodically analyzes historical collision processing records, compares the intelligent filtering results with the results of manual confirmation by designers based on professional knowledge and CAD models, optimizes rule weights through reinforcement learning mechanisms, and uses a reward function to evaluate the consistency between machine judgment and manual confirmation results.

[0117] Through iterative updates, the confidence level of the filtering rules is dynamically adjusted, enabling the platform's intelligent filtering capabilities to continuously accumulate into unique CAD design knowledge for enterprises along with projects.

[0118] S501: The adaptive optimization unit periodically runs its background learning process. Its core data source for learning is the machine's preliminary judgment generated by the intelligent collision screening module (S202) and the final human confirmation given by the designer after final verification within a historical period. These data are expert feedback datasets, stored in a dedicated optimization log table, with the following structure: unique collision identifier, ID of the involved 3D model component, machine judgment, specific rule ID of the application, final human confirmation, and confirmation time. Based on the data pairs, a reinforcement learning reward logic matrix table as shown in Table 2 is constructed.

[0119] In this embodiment, the comparison results are divided into four categories: If the machine exempts the user's application and the user ignores it, it is recorded as a correct exemption. If the machine pushes the information and a human ultimately resolves it, it is recorded as a correct identification. The most dangerous error is one that is exempted by the machine but ultimately marked as resolved by the human, and is recorded as a missed filter or a false negative, meaning that a real conflict was incorrectly filtered out by the system. If a message is pushed by the machine but is ultimately marked as ignored by a human, it is considered an efficiency error and is recorded as a false alarm or a false positive.

[0120] Table 2. Reinforcement Learning Reward Logic Matrix

[0121] In Table 2, α1 is the positive reward coefficient for correct exemption, β3 is the heavy penalty coefficient for missed detection, β2 is the light penalty coefficient for false positives, and α2 is the positive reward coefficient for correct identification. These coefficients correspond to the global periodic reward function R in subsequent S502. total The parameters a, b, and c in the matrix correspond as follows: a = α1 = 1, b = |β3| = 10, c = |β2| = 2, d = α2 = 1. Using this matrix, the reinforcement learning algorithm rewards or penalizes the machine's judgment based on the final human confirmation, thereby updating the rule confidence.

[0122] S502: A Bayesian optimization-guided policy gradient reinforcement learning framework is used to dynamically adjust and optimize the confidence threshold of each rule in the intelligent collision filtering knowledge base and the dynamic decision threshold parameter in S303.

[0123] To achieve this functionality, a dedicated feature vector is established to represent the application scenario of each rule, such as rule type, involved component professional combination, spatial region, historical accuracy, etc. The agent of the reinforcement learning algorithm is a set of these parameters, and its goal is to maximize a global periodic reward function based on the safety and efficiency trade-off definition of power generation engineering. ; in, This represents the total number of active rules in the knowledge base. Corresponding rules The number of correct exemptions, correct identifications, missed detections, and false alarms generated during this learning cycle; a, b, c, and d are preset reward and penalty coefficients, typically set as a=1, b=10, c=2, d=1, corresponding to the coefficients in Table 2 (b and c are the absolute values ​​of the penalty coefficients): for missed detections... Impose the most severe punishment. The maximum value indicates a high likelihood of false alarms. Apply moderate penalties to balance efficiency, while providing appropriate rewards for correct identification and exemption.

[0124] in, To match the S303 dynamic decision threshold The relevant performance rewards are used to incorporate the flow decision strategy into reinforcement learning optimization. The measure is the proportion of models that are allowed through the current threshold strategy and subsequently receive good feedback. For example, it involves statistically analyzing all models that have passed the threshold in the most recent evaluation period. The proportion of models that were deemed suitable for release without requiring significant rework, design changes, or new valid collisions during downstream professional or construction phases. This is a weighting factor used to adjust the relative importance between collision screening rewards and threshold performance rewards; its value is the same as described above. , , The coefficients are independent of each other and can be set according to the actual needs of the project. By introducing this factor, the learning process not only optimizes the collision filtering rules, but also optimizes the entry threshold for model release, enabling the decision points of the entire system to evolve collaboratively.

[0125] Specifically, at the end of each learning cycle, the algorithm will evaluate the collected feedback and the calculated reward. Update model parameters (including weighting coefficients, decision rules, etc.). For example, if a rule regarding the safe distance between pipes and cable trays is frequently generated... The confidence level will be significantly lowered, making it more likely that similar collisions will be marked as push notifications for manual review in the future.

[0126] Furthermore, a long-term, stable rule for the overlap of duct and structural pre-reserved opening edges will gain higher confidence, enabling it to more decisively filter out more such compliance collisions. Meanwhile, for specific project phases, if the overall model quality tends to stabilize, it may be recommended to lower γ to adjust the decision threshold. It is more flexible to improve collaborative workflow efficiency. To ensure the safety and controllability of the learning process, the optimization engine has a sandbox mode: all parameter adjustments are first evaluated in an isolated simulation environment built from historical data. Only when the new strategy meets the preset safety and performance indicators in the simulation will it be approved for use in the production environment.

[0127] In addition, all parameter adjustment records, reward and punishment function calculation results, and simulation evaluation reports are stored in the database table for system administrators to audit and manually intervene, such as manually freezing the adjustment of a certain rule.

[0128] Furthermore, through this continuous, closed-loop adaptive optimization, the core intelligence of this embodiment is reflected not only in the precise collision filtering of S200 but also in the dynamic decision-making of S300. It is not static but evolves continuously with differences in specific project types, design stages, and design habits. Most importantly, it systematically and quantifiably encodes the professional judgment and experience in handling collisions and solidifies it into the platform's knowledge base and decision parameters, making it a digital asset for the enterprise.

[0129] This embodiment provides a collaborative management and control method for 3D design in power generation engineering. Through S100, a source-controlled and quantifiable data benchmark is established, providing a reliable and unified 3D CAD model data source and project management benchmark for the entire collaborative design process. Through S200, intelligent pre-screening based on power generation engineering domain knowledge and 3D model semantics is applied to efficiently separate noise and signals from massive amounts of raw geometric collisions, directly impacting the quality control of 3D design results. Through S300, a multi-dimensional state assessment integrating model geometric robustness, version standardization, attribute completeness, and operational risks is introduced, combined with collision resolution rate and dynamic decision thresholds for global project status, forming an adaptive and unavoidable rigid quality gate for 3D CAD model release. Through S400, automated process triggering based on macroscopic 3D model state indicators is implemented, improving cross-disciplinary collaboration efficiency. Finally, through S500, a self-optimization capability centered on reinforcement learning is formed, enabling intelligent rules to continuously learn and evolve from historical 3D model processing and human feedback, improving design efficiency and knowledge accumulation based on 3D models while ensuring engineering quality and safety.

[0130] To verify the superiority of the method in this embodiment, the following comparative experiment was designed: A typical 660MW ultra-supercritical unit main plant 3D design model was selected as the test object, covering mechanical, electrical, thermal control, civil engineering and other disciplines, including components such as pipelines, air ducts, cable trays, equipment foundations, etc., with a total number of no less than 5000 components. The test set was constructed from historical collision records, in which real collisions (which need to be resolved manually) and compliant collisions (which should be filtered) were mixed in actual proportions, and the methods shown in Table 3 were selected for comparative experiments.

[0131] Table 3. Comparison Methods

[0132] The evaluation indicators are as follows: False alarm rate: The number of collisions that were still pushed after rule filtering but manually judged as "compliant and ignored" / total number of push collisions × 100%; Collision miss rate: (Number of collisions automatically filtered by the machine but manually judged as "needing to be resolved") / Total actual collisions × 100%; Average effective collision handling time: The average time (in hours) from collision notification to manual confirmation of closed loop; Model transfer rejection rate: The proportion of model transfer applications rejected by logic gates; Downstream rework rate: The proportion of release models that require rework during downstream professional or construction phases.

[0133] Figure 4A bar chart comparing the performance of different methods (the horizontal axis represents the method number, and the vertical axis represents the percentage of the corresponding indicator / time). Figure 4 This can be seen intuitively: The complete solution A has a false alarm rate of only 8%, which is much lower than C (45%) which lacks stage 2 collision recognition and the existing technology E (50%), proving that engineering rule semantic filtering can significantly reduce false alarms; The false alarm rate of A is 2%, which is slightly higher than that of pure geometric detection C (0%). However, considering that the false alarm rate of C is as high as 45%, its actual effective collision handling efficiency is far lower than that of A. A has the shortest average processing time (2.5 hours), and its flow rejection rate (12%) and downstream rework rate (4%) are significantly better than other schemes, indicating that multidimensional credibility assessment and dynamic threshold rigid logic gate can effectively intercept low-quality models and reduce downstream rework. Schemes B and D, which lack any dimension, deteriorated in all indicators, verifying the necessity of integrating all dimensions in this application.

[0134] The above comparative experiments fully demonstrate that this embodiment achieves an optimal balance between security and efficiency through the synergistic effect of three-dimensional monitoring of data homogenization health index, phase two engineering semantic filtering, four-dimensional evaluation of model comprehensive credibility, and dynamic adaptive threshold rigid logic gates, and has significant technical progress.

[0135] This embodiment provides a collaborative management and control method for three-dimensional design of power generation projects. By connecting with authoritative project data sources through data homogenization and rule configuration units, and adopting a three-level coding system of project-volume-component, the problem of data heterogeneity is eliminated from the source. At the same time, through the professional rule library built into the intelligent collision recognition engine, compliance false collisions are automatically filtered out, and design resources are concentrated on solving real design conflicts.

[0136] This embodiment provides a collaborative management and control method for three-dimensional design of power generation projects. It uses a dynamic correlation verification and control unit to calculate and evaluate the model requesting transfer. This evaluation model integrates the model's multi-dimensional technical status reliability. Effective collision resolution rate And dynamically adjust the decision threshold based on the overall status of the project. The process can only continue when both hard indicators are met simultaneously, forming a technology-driven and unavoidable quality control closed loop.

[0137] This embodiment provides a collaborative management and control method for three-dimensional design of power generation projects. Through a progress and resolution rate monitoring mechanism, it realizes the automatic creation and assignment of collaborative tasks driven by project status, changes passive waiting instructions to active task pushing, and improves the speed of collaborative response.

[0138] This embodiment provides a collaborative management and control method for three-dimensional design of power generation projects. Based on historical processing data, it dynamically optimizes filtering rules and parameters through reinforcement learning, enabling the platform to evolve itself as project experience accumulates.

[0139] Example 2 This embodiment provides a collaborative management and control system for three-dimensional design of power generation projects, including: Data homogenization and rule configuration module: responsible for establishing authoritative data benchmarks and configuring various business rules.

[0140] Collaborative Execution and Status Acquisition Module: Serves as the unified entry point for user operations and collects data throughout the entire process. Dynamic Association Verification and Control Module: Performs mandatory quantitative assessments and workflow decisions. Global Visual Management Module: Provides visualized monitoring and penetrating query capabilities for the project's macro-level status. Adaptive Optimization Module: Optimizes platform rules based on historical feedback data through reinforcement learning mechanisms. Platform Data Center: Stores all structured and unstructured data based on distributed relational and non-relational databases.

[0141] This embodiment provides a collaborative management and control system for 3D design of power generation projects, implemented using a distributed microservice architecture deployed on a server cluster. The data homogenization and rule configuration module serves as a configuration center service, providing a graphical interface for administrators to maintain upstream system connection configurations, WBS mapping relationships, and the project rule base. The collaborative execution and status acquisition module includes a set of adapter microservices responsible for integration with different CAD / BIM software, a file storage and lightweight conversion service, and a user portal web application. The core of the dynamic association verification and control module is a high-performance computing engine service responsible for real-time execution of complex evaluation algorithms S301-S305. The global visualization management module is built based on a WebGL (3D drawing protocol) 3D rendering engine and real-time data stream service. The adaptive optimization module is an independent machine learning scheduling service that extracts data from the platform data center according to a plan and runs training tasks. The platform data center stores structured business data based on a distributed relational database, 3D model files based on object storage, and performance indicators and logs using a time-series database. Services communicate loosely through message queues.

[0142] It should be noted that each module in this embodiment corresponds one-to-one with each step in Embodiment 1, and their specific implementation processes are the same, so they will not be repeated here.

[0143] Example 3 This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the collaborative management and control method for three-dimensional design of power generation projects as described in Embodiment 1 above.

[0144] Example 4 This embodiment provides a computer device, such as... Figure 5 As shown, the system includes a computer-readable storage medium 1003, a processor 1001, a communication interface 1002, and a computer program stored on the computer-readable storage medium 1003 and executable on the processor 1001. The processor 1001, communication interface 1002, and computer-readable storage medium 1003 can be connected via a bus or other means. The communication interface 1002 is used to receive and send data. When the processor 1001 executes the program, it implements the steps in the collaborative control method for three-dimensional design of power generation projects as described in Embodiment 1 above.

[0145] When the processor 1001 executes the steps in the collaborative management and control method for three-dimensional design of power generation engineering as described in Embodiment 1, it can drive the computer equipment to perform a series of functions, including API-level deep data interaction with industrial three-dimensional CAD / BIM software, parsing and processing three-dimensional model geometry and attribute information, performing intelligent collision screening based on engineering semantics, implementing multi-dimensional model technical status assessment, generating dynamic decision thresholds, performing mandatory rigid release judgments, and realizing adaptive optimization of the knowledge base based on reinforcement learning, and accurately executes the specific algorithms, formulas and logical judgments in each step.

[0146] The computer device provided in this embodiment can be a physical server or virtual machine with the aforementioned software deployed, or a client workstation that accesses the platform service via a network. In the client workstation scenario, it is equipped with a professional CAD / BIM plugin that is deeply integrated with the platform. This plugin embeds local geometric preprocessing, operation acquisition, and communication modules for efficiently uploading models and receiving platform instructions.

[0147] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A collaborative management and control method for three-dimensional design of power generation projects, characterized in that, include: A two-way data channel is established with the upper-level design system or the lower-level design database to synchronize project data to the local machine. During the data synchronization process, the data homogeneity health index is monitored to determine whether the data protection protocol is triggered. The data homogeneity health index integrates data structure consistency, content synchronization timeliness, and data source availability. After completing the project data synchronization, in response to the collision check command, all triangular surface patches in the model to be checked are inspected to form an original collision list. Each collision item in the original collision list is traversed, and the corresponding judgment rules in the power generation engineering professional rule knowledge base are matched and executed. The collision items in the original collision list are filtered to obtain valid collision items and push them to complete the collision rectification. In response to the model transfer instruction, a comprehensive evaluation of the model to be transferred is performed to obtain the comprehensive credibility value of the current technical status of the model. Valid collision items and collision rectification results associated with the model to be transferred are retrieved, the resolution progress rate is calculated, and the comprehensive credibility value of the current technical status of the model and the resolution progress rate are integrated to give a decision on whether to allow or reject the transfer.

2. The collaborative management and control method for three-dimensional design of power generation projects as described in claim 1, characterized in that, Also includes: The valid collision items pushed are compared with the results of manual verification to obtain the comparison results. The global periodic reward function is calculated, and parameters and rules are adjusted in combination with reinforcement learning algorithms.

3. The collaborative management and control method for three-dimensional design of power generation projects as described in claim 1, characterized in that, The overall reliability value of the model's current technical state is: ； in, The geometric robustness score is obtained by weighting the results after counting the number of pass, warning, and failure items through model closure checks, invalid geometry checks, and normal consistency checks. The attribute information completeness score is obtained by statistically analyzing the fill rate of required attribute fields for all components in model k; operational risk entropy. M represents the total number of preset operation types. For the first observation window The proportion of the number of occurrences of a particular type of operation out of the total number of operations; , , ,and Weighting coefficients; Version evolution normativity score =(Stability Score × 0.4 + Annotation Quality Score × 0.4 + (1 - Rollback Ratio) × 0.2); where the stability score is calculated as follows: extract the timestamp sequence of the most recent N model k submissions, and calculate the standard deviation of the interval between two adjacent submissions. Compared to average submission interval The ratio of these ratios is used to obtain the stability score through a normalized mapping. The calculation logic for the annotation quality score is as follows: a thesaurus containing four types of required keywords, namely, modification location, reason for change, associated volume, and collision number, is preset. The percentage of annotations that hit all required keywords in the most recent N submissions is counted and directly used as the annotation quality score.

4. The collaborative management and control method for three-dimensional design of power generation projects as described in claim 1, characterized in that, The resolution progress rate is: ； in, The total number of collisions that have been closed-looped among all valid collision items with component associations in model k; This represents the total number of all valid collisions that have a component relationship with model k.

5. The collaborative management and control method for three-dimensional design of power generation projects as described in claim 1, characterized in that, The specific steps for making a decision on whether to approve or reject the application are as follows: The overall credibility value of the current technical state of the model based on each model k. Calculate the average credibility of all uploaded models in the current project. And based on average confidence level Generate dynamic decision thresholds ;in, As a static base threshold, This is the negative feedback adjustment coefficient. This is the historical average credibility benchmark for the project; Based on the overall credibility value of the model's current technological state and progress rate The comprehensive intelligent decision-making score published by the computational model ;in, and These are the weighting coefficients; The decision to approve or reject a request is determined by a rigid and unavoidable logic gate: ;in, For logical AND.

6. The collaborative management and control method for three-dimensional design of power generation projects as described in claim 1, characterized in that, The data homogenization health index is: ； in, To ensure data structure consistency, a weighted average of the similarity between the local copy of each data source at time point t and the authoritative copy of the data source is calculated. To ensure timely synchronization of content; To ensure data source availability, the validity of status codes and Secure Sockets Layer certificates is combined. To penalize the sensitivity coefficient, Penalty for abnormal frequency; , and These are the weighting coefficients.

7. The collaborative management and control method for three-dimensional design of power generation projects as described in claim 6, characterized in that, The calculation steps for the frequency anomaly penalty are as follows: A preset synchronization period table stores the preset synchronization interval rules for all data sources; Build a log table of actual synchronization attempts to record the execution information of each synchronization task in full; Within the statistical period, the preset synchronization period table is traversed. For each data source, the total number of synchronizations that should be completed within the statistical period is calculated. That is, the theoretical number of synchronizations for a single data source = total duration of the statistical period / preset synchronization interval of the data source. The theoretical number of synchronizations for all data sources is summed to obtain the total theoretical number of synchronizations. ; Filter the records in the actual synchronization attempt log table that fall within the current statistical period and are marked as successful, and sum them up to get the total number of successful synchronizations across all data sources. ; Calculate frequency anomaly penalty .

8. A three-dimensional design collaborative management and control system for power generation projects, characterized in that, include: The data homogenization and rule configuration module is configured to: establish a two-way data channel with the upper-level design system or the lower-level design database, synchronize project data to the local machine, and monitor the data homogenization health index during the project data synchronization process to determine whether the data protection protocol is triggered; the data homogenization health index integrates data structure consistency, content synchronization timeliness, and data source availability; The collaborative execution and status acquisition module is configured to: after completing the project data synchronization, respond to the collision check command, check all the triangular surface patches in the model to be checked, form an original collision list, traverse each collision item in the original collision list, match and execute the corresponding judgment rules in the power generation engineering professional rule knowledge base, filter the collision items in the original collision list, obtain the valid collision items and push them to complete the collision rectification. The dynamic association verification and control module is configured to: respond to the model transfer instruction, comprehensively evaluate the model to be transferred, obtain the comprehensive credibility value of the current technical status of the model, retrieve the valid collision items and collision rectification results associated with the model to be transferred, calculate the resolution progress rate, integrate the comprehensive credibility value of the current technical status of the model and the resolution progress rate, and give a decision to allow or reject the transfer.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the collaborative management and control method for three-dimensional design of power generation engineering as described in any one of claims 1-7.

10. A computer device comprising a computer-readable storage medium, a processor, and a computer program stored on the computer-readable storage medium and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the collaborative management and control method for three-dimensional design of power generation engineering as described in any one of claims 1-7.

Images