Automatic examination and scoring system for completion delivery model based on large model
By using an automated review and scoring system based on large models, the problems of reliance on manual data entry and low efficiency in updating the rule base in existing technologies have been solved. This system enables efficient and accurate automated review and scoring of BIM models, improving the compliance of models and the relevance of scoring.
Patent Information
- Application Number
- CN202511091107.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-05
- Publication Date
- 2025-11-18
AI Technical Summary
Existing automated review platforms rely on manual data entry, making it difficult to adapt to frequently updated local policies. Their rule base updates are inefficient, their review granularity is coarse, they cannot parse the implicit logic in standard texts, their scoring performance is poor, and they lack targeted guidance.
An automated review and scoring system for completed delivery models based on large models is adopted, including a review standard library and dynamic update module, a BIM model review module, an automatic scoring module, a report generation and interaction improvement module, and a knowledge management module. It uses a multimodal large model to parse review standard documents, builds a cross-regional review standard library through transfer learning, and automates BIM model review and generates scoring reports.
It enables efficient and accurate automatic review of BIM models, improves model compliance and accuracy, provides timely improvement suggestions, and enhances the adaptability of the rule base and the relevance of the scoring.
Smart Images
Figure CN120975733A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of artificial intelligence technology, and in particular to an automatic review and scoring system for as-built delivery models based on large models. Background Technology
[0002] With the promotion and development of BIM technology, the requirements for the inspection and review of BIM models are increasing. Many provinces and cities across China have introduced their own BIM model standards. Some regions have also launched platforms for automatic model review. However, existing automatic review platforms have significant shortcomings. Firstly, the parsing of existing standard texts relies on manual input, making it difficult to adapt to frequently updated local policies. Secondly, the rule base update efficiency is low. Thirdly, the review granularity is coarse; existing systems can only check explicit parameters, such as numerical values and classification codes, and cannot parse the implicit logic in the standard text, resulting in poor scoring and a lack of targeted and actionable guidance. Summary of the Invention
[0003] The purpose of this invention is to provide an automatic review and scoring system for as-built delivery models based on large models, in order to solve the problems mentioned in the background art.
[0004] To achieve the above objectives, this invention provides an automatic review and scoring system for as-built delivery models based on a large-scale model. The system includes a review standard library and dynamic update module, a BIM model review module, an automatic scoring module, a report generation and interactive improvement module, and a knowledge management module. The review standard library and dynamic update module constructs and updates the review standard library based on a multimodal large-scale model and transfer learning. The BIM model review module reviews BIM models based on the review standard library. The automatic scoring module scores models based on the review results. The report generation and interactive improvement module generates a review report based on the review and scoring results and provides improvement suggestions. The knowledge management module manages the project review data.
[0005] Preferably, the steps for constructing the review standard library and the dynamic update module include:
[0006] Obtain the BIM model review standard document, parse the review standard document using a multimodal large model, and extract the rules in the review standard document, including structured rules, unstructured rules, naming rules, and parameter constraints;
[0007] Based on the extraction rules, a rule knowledge graph is constructed by extracting entity relationships.
[0008] After the logical association is completed, a cross-regional review standard library is established through transfer learning, conflicting clauses are aligned, and dynamic adaptation rules are generated.
[0009] Preferably, the structured rules include component naming conventions, LOD level requirements, and parameter thresholds.
[0010] Preferably, updating the review standards library includes: real-time monitoring of review standards document updates, and incremental learning to update the review standards library after a new review standards document is published.
[0011] Preferably, the BIM model review module reviews the BIM model based on a review standard library, including:
[0012] Import the BIM model into the BIM model review module and select project information; the module will automatically match the corresponding BIM standards in the review standard library.
[0013] By combining geographical location and enterprise information, the system selects the standard combination that best meets the project requirements and automatically loads the corresponding review rules, and reviews the BIM model according to the review rules.
[0014] Preferably, the review rules include geometric information review, parameter information review, component specification review, collision detection, and model integrity review, wherein collision detection refers to detecting collisions between structures and equipment.
[0015] Preferably, the automatic scoring module automatically scores the model based on the review results, including:
[0016] Customize rating priorities and importance, and use machine learning models to dynamically adjust rating weights based on the defined rating priorities and importance;
[0017] The automatic scoring module scores the BIM model based on the review results and the adjusted scoring weights. The scoring criteria are based on preset rules in the review standard library. The scoring items include model completeness, model accuracy, and standard compliance. Model accuracy includes the accuracy of geometric and parametric information, and standard compliance refers to whether the component naming and classification are standardized.
[0018] Preferably, the review report in the report generation and interaction improvement module includes detailed scoring information for each review item, specific components that failed the review and their corresponding problem descriptions, and specific improvement suggestions for the failed items.
[0019] Preferably, the knowledge management module includes:
[0020] Project storage unit: Stores review results and scoring results for multiple BIM models;
[0021] Project Search Unit: By constructing an industry knowledge graph, linking standard clauses, review cases and remediation solutions, historical BIM models can be retrieved.
[0022] Project Management Unit: Builds a dedicated enterprise review knowledge base, automatically archives frequently asked questions and solutions, and generates standard trend reports.
[0023] Therefore, the above-mentioned automatic review and scoring system for as-built delivery models based on large models, as described in this invention, has the following beneficial effects:
[0024] (1) By parsing the review standard documents through a multimodal large model, the implicit logic in the review standard documents can be extracted, and the limitations of the traditional manual entry of rule base are broken.
[0025] (2) The BIM model review module automatically reviews the BIM model, which improves the review efficiency of the BIM model and avoids the time-consuming and error-prone manual inspection. At the same time, the system can automatically score based on the review results and provide timely improvement suggestions to ensure the compliance and accuracy of the model.
[0026] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0027] Figure 1 This is a system structure diagram of an embodiment of the present invention;
[0028] Figure 2 This is a flowchart illustrating the construction process of the review standard library for embodiments of the present invention. Detailed Implementation
[0029] 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. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. In the description of the present invention, it should be noted that the terms "upper," "lower," "inner," "outer," etc., indicating orientation or positional relationships are based on the orientation or positional relationships shown in the accompanying drawings, or the orientation or positional relationships commonly used when the product of the invention is in use. They are only for the convenience of describing the present invention and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention.
[0030] Example
[0031] This invention provides an automatic review and scoring system for as-built delivery models based on large models, including a review standard library and dynamic update module, a BIM model review module, an automatic scoring module, a report generation and interaction improvement module, and a knowledge management module.
[0032] Review Standard Library and Dynamic Update Module: Builds and updates the review standard library based on multimodal large models and transfer learning.
[0033] Building a review standards library includes:
[0034] First, obtain BIM model review standard documents from different countries, regions, and enterprises. Then, using the natural language processing capabilities of multimodal large models, automatically parse the BIM model review standard documents from different countries, regions, and enterprises, and extract the rules from the review standard documents.
[0035] For BIM model review standard documents from different countries and regions, extract structured and unstructured rules. Structured rules include component naming conventions (such as "wall_concrete_200mm"), LOD level requirements, and parameter thresholds. Unstructured rules are implicit rules in the BIM model review standard documents, such as "green buildings must label carbon emission parameters".
[0036] For the enterprise's BIM model review standard documents, extract the naming rules and parameter constraints.
[0037] Secondly, based on the extraction rules, a rule knowledge graph is constructed by extracting entity relationships to obtain logical associations, such as "fire protection facilities must be associated with material parameters".
[0038] After logical association is completed, a cross-regional review standard library is established through transfer learning to align conflicting clauses. For example, Beijing requires "fire door fire resistance time ≥ 1.5h" while Guangdong requires "≥ 1.2h," generating dynamic adaptation rules. The review standard library adopts a modular design and can be categorized according to different BIM standards, including: ① National-level BIM specifications (such as the "Building Information Modeling Design and Delivery Standard"). ② Local BIM standards from different regions. ③ Enterprise-customized BIM specifications (such as component naming rules, model detail level requirements, etc.).
[0039] Finally, the system monitors for updates to review standard documents in real time, and updates the review standard library through incremental learning when new review standard documents are released. When a new standard is detected, a discrepancy report is automatically generated and the review logic is updated.
[0040] BIM Model Review Module: Reviews BIM models based on a review standards library. Specifically:
[0041] Import the BIM model into the BIM model review module using the Revit plugin and select project information; the module will automatically match the corresponding BIM standards in the review standards library.
[0042] The Revit plugin combines geographic location and company information to select the standard combination that best meets project requirements and automatically loads the corresponding review rules. The BIM model is then reviewed according to these rules. Review rules include geometric information review, parametric information review, component specification review, clash detection, and model integrity review.
[0043] Among them, geometric information review: to check whether the geometric accuracy of the components in the model meets the standard requirements, especially whether the size, shape, spacing and other aspects of the components meet the design drawings.
[0044] Parameter Information Review: Inspect the parameter settings in components, such as material, name, classification, and other information, to ensure they comply with the requirements of the standard library. For example, whether the component name follows the standard format, and whether the material parameters are set correctly. The specific review method is as follows: Based on the parsed rule base, check whether the BIM model component parameters conform to the standards, such as material, fire resistance rating, and classification code; verify the parameter correlation through logical reasoning, such as whether "fire door" is associated with the "fire resistance rating" parameter.
[0045] Component specification review: Compare the component classification in the model with the specifications in the standard library to check whether the components are classified according to BIM standards and verify whether the type of each component meets the requirements. This includes: Fuzzy matching and error correction: Identify near-name errors (e.g., "drainage pipe" incorrectly labeled as "sewage pipe"); Naming recommendation: For unnamed or non-standard named components, automatically generate standard-compliant names (e.g., "WALL_01" corrected to "exterior wall_concrete_300mm").
[0046] Collision detection: Automatically detects collisions in the model and provides reasonable suggestions based on standards.
[0047] Model integrity review: Check whether the model's annotations, drawing accuracy, etc., comply with relevant specifications, and whether it contains all the necessary components and information for all design stages.
[0048] It is important to note that during the review process, when multiple standards conflict, such as enterprise standards and local standards, the review basis is automatically selected based on priority rules (such as "local standards take precedence"), and a conflict explanation document is generated.
[0049] Automatic scoring module: Scores the BIM model based on the review results, specifically:
[0050] First, define the rating priorities and importance. Then, use a machine learning model to dynamically adjust the rating weights based on the defined rating priorities and importance.
[0051] Finally, the automatic scoring module scores the BIM model based on the review results and the adjusted scoring weights. The scoring results are displayed in the form of charts. The scoring criteria are based on the preset rules in the review standard library. The scoring items include model completeness, model accuracy, and standard compliance. Model accuracy includes the accuracy of geometric information and parametric information, and standard compliance refers to whether the component naming and classification are standardized.
[0052] The report generation and interaction improvement module generates an NLP review report based on the review and scoring results, and provides improvement suggestions. The NLP review report includes detailed scoring information for each review item, the specific components that failed the review and their corresponding problem descriptions, and specific improvement suggestions for the failed items, such as "Component FM-001: Fire resistance time is missing, violating Article 5.2 of Beijing 2024 Standard". Improvement suggestions include providing actionable text commands such as "Add a 'Fire Resistance Time' field to the Properties panel and set it to 1.5h".
[0053] User feedback loop: Allow users to mark cases of misjudgment, such as "a certain component has a parameter deviation allowed due to special process", and add them to the enterprise's exclusive review knowledge base to optimize the review logic of subsequent projects.
[0054] Knowledge Management Module: Manages project review data, specifically including:
[0055] Project storage unit: Stores review results and scoring results of multiple BIM models. Under the premise of privacy protection, it improves the rule generalization ability by accumulating automatic review data results of multiple projects.
[0056] Project Search Unit: By constructing an industry knowledge graph, linking standard clauses, reviewing cases and remediation solutions, historical BIM models can be retrieved, such as searching for "curved curtain wall error processing".
[0057] Project Management Unit: Builds a dedicated enterprise review knowledge base, automatically archives frequently asked questions and solutions, such as "the allowable parameter tolerance for a certain project is ±5%", and generates standard trend reports.
[0058] Therefore, the present invention adopts the above-mentioned automatic review and scoring system based on a large model of completed delivery, which, by integrating machine learning, natural language processing and transfer learning technologies, can extract the implicit logic in the review standard documents, align with regional differences and improve the accuracy of scoring.
[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. A large model-based as-built delivery model automatic review scoring system, characterized in that: The BIM model review module reviews the BIM model based on the review standard library; the automatic scoring module scores according to the review result; the report generation and interactive improvement module generates a review report according to the review and scoring results, and gives improvement suggestions; The knowledge management module manages project review data. The review standard library and dynamic updating module includes the following steps of constructing the review standard library:
2. The large model-based completion delivery model automatic review scoring system according to claim 1, wherein, Obtain the BIM model review standard document, use the multi-modal large model to analyze the review standard document, and extract the rules in the review standard document, including structured rules, unstructured rules, naming rules and parameter constraints; According to the extracted rules, the rule knowledge graph is constructed through entity relationship extraction; After logical association, cross-regional review standard library is established through transfer learning, conflict clauses are aligned, and dynamic adaptive rules are generated. The structured rules include component naming specification, LOD level requirement and parameter threshold.
3. The large model-based as-built delivery model automatic review scoring system of claim 2, wherein: The update of the review standard library includes real-time monitoring of the update of the review standard document, and updating the review standard library through incremental learning after the new review standard document is published.
4. The large model-based completion delivery model automatic review scoring system of claim 1, wherein: The BIM model review module reviews the BIM model based on the review standard library includes:
5. The large model-based completion delivery model automatic review scoring system of claim 1, wherein, Import the BIM model into the BIM model review module and select the project information, automatically match the corresponding BIM standard in the review standard library; Select the standard combination that best meets the project requirements according to the geographical location and enterprise information, and automatically load the corresponding review rules to review the BIM model according to the review rules. The review rules include geometric information review, parameter information review, component specification review, collision detection and model integrity review, wherein the collision detection refers to the detection of the collision between structures and equipment.
6. The large model-based as-built delivery model automatic review scoring system of claim 5, wherein: The automatic scoring module automatically scores the model according to the review result includes:
7. The large model-based completion delivery model automatic review scoring system of claim 1, wherein: Customize the scoring priority and importance, and dynamically adjust the scoring weight according to the defined scoring priority and importance by using machine learning model; The automatic scoring module scores the BIM model based on the review result and the adjusted scoring weight, the scoring standard is based on the preset rules in the review standard library, and the scoring items include model integrity, model accuracy and standard compliance, wherein the model accuracy includes geometric information and parameter information accuracy, and the standard compliance is whether the component naming and classification are standardized. The review report in the report generation and interactive improvement module includes the detailed scoring situation of each review item, the specific components that do not pass the review and their corresponding problem description, and the specific improvement suggestions for the project that does not pass the review.
8. The large model-based completion delivery model automatic review scoring system of claim 1, wherein: The knowledge management module includes:
9. The large model-based completion delivery model automatic review scoring system of claim 1, wherein, Project storage unit: store multiple BIM model review results and scoring results; Project search unit: through the construction of industry knowledge graph, associate standard clauses, review cases and repair schemes, and search historical BIM models. Project Management Unit: Build enterprise-specific review knowledge base, automatically archive high-frequency issues and solutions, and generate standard trend reports.