Building information model examination method and system based on large language model driven multi-Agent workflow

Abstract

The invention discloses a building information model examination method and system based on large language model driven multi-Agent workflow, and the method comprises the steps: obtaining a to-be-examined building information model and all CAD drawing data; processing all to-be-examined building information text data and all CAD drawing data in the to-be-examined building information model by adopting a large language model technology to obtain building model data and building drawing data, splitting the building model data, and taking the building drawing data as an examination standard of each examination process, and controlling an Agent system matched with each piece of building model sub-data obtained after splitting to inspect each piece of building model sub-data until an inspection result of each piece of building model sub-data meets a preset condition, and outputting a summarized result of all the inspection results as a target inspection result of the to-be-inspected building information model. According to the building information model examination method and system based on the large language model driving multi-Agent workflow provided by the embodiment of the invention, the examination efficiency of the building information model to be examined is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of building information model review, and in particular to a building information model review method and system based on a large language model driven multi-agent workflow. Background Art

[0002] With the rapid development of the construction industry, BIM technology plays an increasingly important role in building design, construction, operation and maintenance. Building Information Model (BIM) is a collection of engineering information resources shared by all stakeholders in the project life cycle, and is also an important basis for all stakeholders to make decisions at different stages.

[0003] Therefore, in order to ensure the validity of model information and improve the quality of model delivery, it is of great practical significance to review the building information model (BIM) before delivery. However, the existing review methods for building information models are inefficient, error-prone, and lack intelligence.

[0004] It can be seen that how to improve the review efficiency of building information models has become a technical problem that technical personnel in this field need to solve urgently. Summary of the invention

[0005] The present invention provides a building information model review method and system based on a large language model driven multi-agent workflow to solve the problems of low efficiency, easy error and lack of intelligence in the existing building information model review method, and improve the review efficiency of the building information model.

[0006] In order to solve the above technical problems, an embodiment of the present invention provides a building information model review method based on a large language model driven multi-agent workflow, which is applied to a building control system architecture constructed according to multi-agent technology.

[0007] Acquire a building information model to be reviewed and various CAD drawing data corresponding to the building information model to be reviewed;

[0008] Using large language model technology to perform structural processing on all the building information text data to be reviewed in the building information model to be reviewed to obtain building model data; and using large language model technology to perform consistency processing on all the CAD drawing data to obtain building drawing data;

[0009] Splitting the building model data, and controlling an Agent system matching each of the building model sub-data obtained after the splitting to review each of the building model sub-data until the review result of each of the building model sub-data meets the preset requirements, wherein each review process is designed to use the building drawing data as a review standard;

[0010] The summary result of all the review results is output as the target review result of the building information model to be reviewed.

[0011] As one of the preferred solutions, the large language model technology is used to perform structured processing on all the text data of the building information to be reviewed in the building information model to be reviewed to obtain the building model data, including:

[0012] Using Revit software to extract each attribute information of each of the building information text data to be reviewed;

[0013] Performing text conversion processing on each of the attribute information to obtain corresponding first attribute text information;

[0014] Performing vector conversion processing on each of the first attribute text information to obtain a corresponding first attribute vector;

[0015] The embedding model is used to process all the first attribute vectors to obtain building model data.

[0016] As one of the preferred solutions, the large language model technology is used to perform uniform processing on all the CAD drawing data to obtain architectural drawing data, including:

[0017] Using a pre-trained image encoder to extract the second attribute text information of each of the CAD drawing data;

[0018] Performing vector conversion processing on each of the second attribute text information to obtain a corresponding second attribute vector;

[0019] The embedding model is used to process all the second attribute vectors to obtain architectural drawing data.

[0020] As one preferred solution, the building model sub-data includes first building model sub-data and second building model sub-data, and the splitting of the building model data includes:

[0021] Splitting the building model data based on data attributes of the building model data to obtain first building model sub-data;

[0022] performing association matching processing on each of the first building model sub-data and the building drawing data to obtain an association matching result for each of the first building model sub-data;

[0023] Each of the first building model sub-data is split based on the association matching result to obtain each second building model sub-data.

[0024] As one preferred solution, the associating and matching processing is performed on each of the first building model sub-data and the building drawing data to obtain the associating and matching result of each of the first building model sub-data, including:

[0025] Acquire a comparison and matching result of each first graphic element based on the similarity between each first graphic element in each first building model sub-data and all second graphic elements in each building drawing data;

[0026] The comparison and matching results of all the first graphic elements are summarized to obtain the associated matching result of each of the first building model sub-data.

[0027] Another embodiment of the present invention provides a building information model review system based on a large language model driven multi-agent workflow, which is applied to a building control system architecture constructed according to multi-agent technology. The system includes a building data acquisition module, a building data processing module, a data splitting and reviewing module, and a review result summary module;

[0028] The building data acquisition module is used to obtain the building information model to be reviewed and each CAD drawing data corresponding to the building information model to be reviewed;

[0029] The building data processing module is used to use the large language model technology to perform structural processing on all the building information text data to be reviewed in the building information model to be reviewed to obtain building model data; and, use the large language model technology to perform consistency processing on all the CAD drawing data to obtain building drawing data;

[0030] The data splitting and reviewing module is used to split the building model data, and control the Agent system matching each building model sub-data obtained after the splitting to review each building model sub-data until the review result of each building model sub-data meets the preset requirements, wherein each review process is designed to use the building drawing data as the review standard;

[0031] The review result summary module is used to output the summary result of all the review results as the target review result of the building information model to be reviewed.

[0032] As one of the preferred solutions, the large language model technology is used to perform structured processing on all the text data of the building information to be reviewed in the building information model to be reviewed to obtain the building model data, including:

[0033] Using Revit software to extract each attribute information of each of the building information text data to be reviewed;

[0034] Performing text conversion processing on each of the attribute information to obtain corresponding first attribute text information;

[0035] Performing vector conversion processing on each of the first attribute text information to obtain a corresponding first attribute vector;

[0036] The embedding model is used to process all the first attribute vectors to obtain building model data.

[0037] As one of the preferred solutions, the large language model technology is used to perform uniform processing on all the CAD drawing data to obtain architectural drawing data, including:

[0038] Using a pre-trained image encoder to extract the second attribute text information of each of the CAD drawing data;

[0039] Performing vector conversion processing on each of the second attribute text information to obtain a corresponding second attribute vector;

[0040] The embedding model is used to process all the second attribute vectors to obtain architectural drawing data.

[0041] As one preferred solution, the building model sub-data includes first building model sub-data and second building model sub-data, and the splitting of the building model data includes:

[0042] Splitting the building model data based on data attributes of the building model data to obtain first building model sub-data;

[0043] performing association matching processing on each of the first building model sub-data and the building drawing data to obtain an association matching result for each of the first building model sub-data;

[0044] Each of the first building model sub-data is split based on the association matching result to obtain each second building model sub-data.

[0045] As one preferred solution, the associating and matching processing is performed on each of the first building model sub-data and the building drawing data to obtain the associating and matching result of each of the first building model sub-data, including:

[0046] Acquire a comparison and matching result of each first graphic element based on the similarity between each first graphic element in each first building model sub-data and all second graphic elements in each building drawing data;

[0047] The comparison and matching results of all the first graphic elements are summarized to obtain the associated matching result of each of the first building model sub-data.

[0048] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0049] (1) In order to facilitate the comparison of the building information text data to be reviewed and the CAD drawing data, so as to realize the review of the building information model to be reviewed, the large language model technology is used to perform structured processing on all the building information text data to be reviewed in the building information model to be reviewed, so as to obtain the building model data; the large language model technology is used to perform consistency processing on all the CAD drawing data, so as to obtain the building drawing data. By using the large language model technology to convert the building information text data to be reviewed and the CAD drawing data into a unified format, the efficiency of the subsequent review of the building information model to be reviewed is improved;

[0050] (2) The building model data is split, and the building drawing data is used as the review standard for each review process. The Agent system that matches the building model sub-data obtained after the split is controlled to review each building model sub-data until the review result of each building model sub-data meets the preset. By dividing the building model data layer by layer into multiple building model sub-data, the corresponding Agent system is used to match and compare each building model sub-data separately, and finally the summary result of all review results is output as the target review result of the building information model to be reviewed, which further improves the review efficiency of the building information model to be reviewed. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flowchart of a building information model review method based on a large language model driven multi-agent workflow in one embodiment of the present invention;

[0052] Figure 2 This is a structural block diagram of a building information model review system based on a large language model driven multi-agent workflow in one embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] In the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0055] In the description of the present application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are only for illustrative purposes, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0056] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0057] An embodiment of the present invention provides a building information model review method based on a large language model driven multi-agent workflow, which is applied to a building control system architecture constructed based on multi-agent technology. For details, see Figure 1 , Figure 1 The flowchart of a building information model review method based on a large language model driven multi-agent workflow in one embodiment of the present invention is shown, which includes steps S1-S4:

[0058] Step S1, obtaining a building information model to be reviewed and various CAD drawing data corresponding to the building information model to be reviewed;

[0059] Step S2: Using the large language model technology to perform structural processing on all the building information text data to be reviewed in the building information model to be reviewed, so as to obtain building model data; and using the large language model technology to perform consistency processing on all CAD drawing data, so as to obtain building drawing data;

[0060] Step S3, splitting the building model data, and controlling the Agent system matching each building model sub-data obtained after the splitting to review each building model sub-data, until the review result of each building model sub-data meets the preset requirements, wherein each review process is designed to use the building drawing data as the review standard;

[0061] Step S4: output the summary result of all review results as the target review result of the building information model to be reviewed.

[0062] Furthermore, in the above embodiment, the large language model technology is used to perform structured processing on all the text data of the building information model to be reviewed to obtain the building model data, including:

[0063] Revit software is used to extract each attribute information of each building information text data to be reviewed, and each attribute information is subjected to text conversion processing to obtain corresponding first attribute text information;

[0064] Each first attribute text information is subjected to vector conversion processing to obtain a corresponding first attribute vector, and all first attribute vectors are processed using an embedding model to obtain building model data.

[0065] It should be noted that the purpose of using the embedding model to process all first attribute vectors to obtain building model data is to capture semantic information through the embedding model, so that similar first attribute text information has similar vector representation.

[0066] Furthermore, in the above embodiment, the large language model technology is used to perform uniform processing on all CAD drawing data to obtain architectural drawing data, including:

[0067] A pre-trained image encoder is used to extract the second attribute text information of each CAD drawing data; each second attribute text information is vectorized to obtain the corresponding second attribute vector; and all second attribute vectors are processed using an embedding model to obtain architectural drawing data.

[0068] Furthermore, in the above embodiment, the building model sub-data includes first building model sub-data and second building model sub-data, and the splitting of the building model data includes:

[0069] Splitting the building model data based on data attributes of the building model data to obtain first building model sub-data;

[0070] Performing association matching processing on each first building model sub-data and the building drawing data to obtain an association matching result for each first building model sub-data;

[0071] Each first building model sub-data is split based on the association matching result to obtain each second building model sub-data.

[0072] It should be noted that the Agent system matching the first building model sub-data can manage the Agent system matching the corresponding second building model sub-data and solve the task with the help of other Agent systems.

[0073] Furthermore, in the above embodiment, the associating and matching processing is performed on each first building model sub-data and the building drawing data to obtain the associating and matching result of each first building model sub-data, including:

[0074] Acquire a comparison and matching result of each first graphic element based on the similarity between each first graphic element in each first building model sub-data and all second graphic elements in each building drawing data;

[0075] The comparison and matching results of all the first graphic elements are summarized to obtain the associated matching result of each first building model sub-data.

[0076] Another embodiment of the present invention provides a building information model review system based on a large language model driven multi-agent workflow, which is applied to a building control system architecture constructed based on multi-agent technology. For details, see Figure 2 , Figure 2 The structure diagram of a building information model review system based on a large language model driving a multi-agent workflow in one embodiment of the present invention is shown. The building information model review system based on a large language model driving a multi-agent workflow includes:

[0077] Building data acquisition module 11: used to obtain the building information model to be reviewed and each CAD drawing data corresponding to the building information model to be reviewed;

[0078] Building data processing module 12: used to perform structural processing on all the building information text data to be reviewed in the building information model to be reviewed by using the large language model technology to obtain building model data; and, to perform consistency processing on all CAD drawing data by using the large language model technology to obtain building drawing data;

[0079] Data splitting and reviewing module 13: used for splitting the building model data, controlling the Agent system matching each building model sub-data obtained after the splitting to review each building model sub-data, until the review result of each building model sub-data meets the preset, wherein each review process is designed to take the building drawing data as the review standard;

[0080] The review result summary module 14 is used to output the summary result of all review results as the target review result of the building information model to be reviewed.

[0081] Furthermore, in the above embodiment, the large language model technology is used to perform structured processing on all the text data of the building information model to be reviewed to obtain the building model data, including:

[0082] Revit software is used to extract each attribute information of each building information text data to be reviewed, and each attribute information is subjected to text conversion processing to obtain corresponding first attribute text information;

[0083] Each first attribute text information is subjected to vector conversion processing to obtain a corresponding first attribute vector, and all first attribute vectors are processed using an embedding model to obtain building model data.

[0084] Furthermore, in the above embodiment, the large language model technology is used to perform uniform processing on all CAD drawing data to obtain architectural drawing data, including:

[0085] A pre-trained image encoder is used to extract the second attribute text information of each CAD drawing data; each second attribute text information is vectorized to obtain the corresponding second attribute vector; and all second attribute vectors are processed using an embedding model to obtain architectural drawing data.

[0086] Furthermore, in the above embodiment, the building model sub-data includes first building model sub-data and second building model sub-data, and the splitting of the building model data includes:

[0087] Splitting the building model data based on data attributes of the building model data to obtain first building model sub-data;

[0088] Performing association matching processing on each first building model sub-data and the building drawing data to obtain an association matching result for each first building model sub-data;

[0089] Each first building model sub-data is split based on the association matching result to obtain each second building model sub-data.

[0090] Furthermore, in the above embodiment, the associating and matching processing is performed on each first building model sub-data and the building drawing data to obtain the associating and matching result of each first building model sub-data, including:

[0091] Acquire a comparison and matching result of each first graphic element based on the similarity between each first graphic element in each first building model sub-data and all second graphic elements in each building drawing data;

[0092] The comparison and matching results of all the first graphic elements are summarized to obtain the associated matching result of each first building model sub-data.

[0093] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A building information model review method based on a large language model driven multi-agent workflow, characterized in that: Applied to a building control system architecture constructed based on multi-agent technology, the method includes: Acquire a building information model to be reviewed and various CAD drawing data corresponding to the building information model to be reviewed; Using large language model technology to perform structural processing on all the building information text data to be reviewed in the building information model to be reviewed to obtain building model data; and using large language model technology to perform consistency processing on all the CAD drawing data to obtain building drawing data; Splitting the building model data, and controlling an Agent system matching each of the building model sub-data obtained after the splitting to review each of the building model sub-data until the review result of each of the building model sub-data meets the preset requirements, wherein each review process is designed to use the building drawing data as a review standard; The summary result of all the review results is output as the target review result of the building information model to be reviewed.

2. The building information model review method based on a large language model driven multi-agent workflow according to claim 1 is characterized in that: The large language model technology is used to perform structural processing on all the text data of the building information to be reviewed in the building information model to be reviewed to obtain building model data, including: Using Revit software to extract each attribute information of each of the building information text data to be reviewed; Performing text conversion processing on each of the attribute information to obtain corresponding first attribute text information; Performing vector conversion processing on each of the first attribute text information to obtain a corresponding first attribute vector; The embedding model is used to process all the first attribute vectors to obtain building model data.

3. The building information model review method based on a large language model driven multi-agent workflow according to claim 1 is characterized in that: The large language model technology is used to perform uniform processing on all the CAD drawing data to obtain architectural drawing data, including: Using a pre-trained image encoder to extract the second attribute text information of each of the CAD drawing data; Performing vector conversion processing on each of the second attribute text information to obtain a corresponding second attribute vector; The embedding model is used to process all the second attribute vectors to obtain architectural drawing data.

4. The building information model review method based on a large language model driven multi-agent workflow according to claim 1, characterized in that: The building model sub-data includes first building model sub-data and second building model sub-data, and splitting the building model data includes: Splitting the building model data based on data attributes of the building model data to obtain first building model sub-data; performing association matching processing on each of the first building model sub-data and the building drawing data to obtain an association matching result for each of the first building model sub-data; Each of the first building model sub-data is split based on the association matching result to obtain each second building model sub-data.

5. The building information model review method based on a large language model driven multi-agent workflow according to claim 4 is characterized in that: The performing association matching processing on each of the first building model sub-data and the building drawing data to obtain an association matching result of each of the first building model sub-data includes: Acquire a comparison and matching result of each first graphic element based on the similarity between each first graphic element in each first building model sub-data and all second graphic elements in each building drawing data; The comparison and matching results of all the first graphic elements are summarized to obtain the associated matching result of each of the first building model sub-data.

6. A building information model review system based on a large language model driven multi-agent workflow, characterized in that: Applied to the building control system architecture formed by multi-agent technology, the system includes a building data acquisition module, a building data processing module, a data splitting and reviewing module and a review result summary module; The building data acquisition module is used to obtain the building information model to be reviewed and each CAD drawing data corresponding to the building information model to be reviewed; The building data processing module is used to use the large language model technology to perform structural processing on all the building information text data to be reviewed in the building information model to be reviewed to obtain building model data; and, use the large language model technology to perform consistency processing on all the CAD drawing data to obtain building drawing data; The data splitting and reviewing module is used to split the building model data, and control the Agent system matching each building model sub-data obtained after the splitting to review each building model sub-data until the review result of each building model sub-data meets the preset requirements, wherein each review process is designed to use the building drawing data as the review standard; The review result summary module is used to output the summary result of all the review results as the target review result of the building information model to be reviewed.

7. The building information model review system based on a large language model driven multi-agent workflow according to claim 6, characterized in that: The large language model technology is used to perform structural processing on all the text data of the building information to be reviewed in the building information model to be reviewed to obtain building model data, including: Using Revit software to extract each attribute information of each of the building information text data to be reviewed; Performing text conversion processing on each of the attribute information to obtain corresponding first attribute text information; Performing vector conversion processing on each of the first attribute text information to obtain a corresponding first attribute vector; The embedding model is used to process all the first attribute vectors to obtain building model data.

8. The building information model review system based on a large language model driven multi-agent workflow according to claim 6, characterized in that: The large language model technology is used to perform uniform processing on all the CAD drawing data to obtain architectural drawing data, including: Using a pre-trained image encoder to extract the second attribute text information of each of the CAD drawing data; Performing vector conversion processing on each of the second attribute text information to obtain a corresponding second attribute vector; The embedding model is used to process all the second attribute vectors to obtain architectural drawing data.

9. The building information model review system based on a large language model driven multi-agent workflow according to claim 6, characterized in that: The building model sub-data includes first building model sub-data and second building model sub-data, and splitting the building model data includes: Splitting the building model data based on data attributes of the building model data to obtain first building model sub-data; performing association matching processing on each of the first building model sub-data and the building drawing data to obtain an association matching result for each of the first building model sub-data; Each of the first building model sub-data is split based on the association matching result to obtain each second building model sub-data.

10. A building information model review system based on a large language model driven multi-agent workflow according to claim 9, characterized in that: The performing association matching processing on each of the first building model sub-data and the building drawing data to obtain an association matching result of each of the first building model sub-data includes: Acquire a comparison and matching result of each first graphic element based on the similarity between each first graphic element in each first building model sub-data and all second graphic elements in each building drawing data; The comparison and matching results of all the first graphic elements are summarized to obtain the associated matching result of each of the first building model sub-data.

Images