Construction quality inspection knowledge base construction method based on natural language extraction and BIM fusion

Abstract

The invention provides a construction quality inspection knowledge base construction method based on natural language extraction and BIM fusion, and the method comprises the steps: firstly obtaining a construction BIM model, extracting BIM information contained in the BIM model through an Ifopen Shell tool, and storing the BIM information as a BIM information table; and then identifying inspection knowledge in the construction quality file by utilizing a gated loop unit model improved based on a deep bidirectional encoder and knowledge feedback in combination with a self-attention mechanism, extracting knowledge information in a text document, and storing the knowledge information as a quality knowledge table. And combining the first information form and the second information form into a BIM-coupled quality knowledge table through word vector semantic similarity matching of three dimensions, combining the BIM information table and quality knowledge into the BIM-coupled quality knowledge table, and constructing a construction quality inspection knowledge base by utilizing an SQL Server. And establishing a virtual-real fusion scene by utilizing Unity, and writing a program to call a knowledge base in the virtual-real scene. The step of knowledge retrieval in quality inspection is simplified, and the efficiency of construction quality inspection is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering construction management, and in particular to a method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration. Background Art

[0002] Project quality inspections are a critical component of construction management, playing a vital role in ensuring an orderly and safe construction site. Currently, quality inspections rely on the experience and knowledge of inspectors, often requiring manual review and retrieval of relevant quality documents to verify that a building meets quality requirements. Manually searching for quality text consumes significant time, and knowledge differences between inspectors can lead to inspection errors. Therefore, simplifying text-based knowledge search methods and using augmented reality (AR) knowledge to guide inspectors during on-site inspections are of great practical value in improving inspection quality and efficiency.

[0003] Natural language processing (NLP) technology can accurately analyze, understand, and process quality document text. Extracting inspection knowledge from quality documents through NLP can effectively reduce the time spent manually searching for quality text. Using BIM as a knowledge carrier and coupling BIM models with quality knowledge can effectively reduce the difficulty of knowledge utilization. Incorporating inspection knowledge into construction scenarios can enhance the efficiency of inspectors and improve inspection quality. Summary of the Invention

[0004] In view of this, the present application provides a method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration, which is used to solve the problems that during the existing construction site inspection, a large amount of quality text needs to be searched and queried, which takes a long time and the query relies on the experience and knowledge of the inspectors.

[0005] In order to achieve the above technical features, the purpose of the present invention is achieved as follows: 1. A method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration, characterized in that the method includes:

[0006] Extract model-related engineering information from construction design drawings and design documents, and add the extracted engineering information to the BIM model using REVIT software;

[0007] Based on the IfcOpenShell library, component entity fields are defined, BIM key semantics are automatically extracted, and stored as a BIM information table to obtain the first information form;

[0008] Based on a deep bidirectional encoder and a gated recurrent unit model improved by knowledge feedback, the self-attention mechanism is combined to identify and extract quality knowledge from quality documents. Regular expressions are used to classify the quality knowledge and store it in a quality knowledge table to obtain the second information form.

[0009] By matching the semantic similarity of word vectors in three dimensions, the first and second information forms are merged into a BIM-coupled quality knowledge table to obtain the third information form;

[0010] Using SQL Server to construct the third information form into a quality inspection knowledge base, thus obtaining a quality inspection database;

[0011] Build a virtual and real scene in Unity, and use VS to write a program to call the inspection knowledge base in the virtual scene.

[0012] Preferably, the engineering information associated with the model is extracted from the construction design drawings and design files, and the extracted engineering information is added to the BIM model based on the REVIT software, including:

[0013] Leveraging computer vision and deep learning technologies, we automatically extract key engineering information from engineering drawings, including the type of project, physical locations, and the relationships and connections between elements.

[0014] Open the building model to which you need to add parameters, select the corresponding component, add the project parameters, and associate the parameters with the element through the SetParameter method of the Element class.

[0015] Preferably, based on the IfcOpenShell library, component entity fields are defined, BIM key semantics are automatically extracted, and stored as a BIM information table to obtain a first information form, including:

[0016] Define key information parameters in the BIM model and convert the BIM model into IFC format file;

[0017] Based on the IfcOpenShell library, a Python program is used to filter the model component entities in the IFC file, read the attribute fields of these component entities, and obtain the key information of each model component;

[0018] The filtered and extracted component key information is sorted and stored as a BIM information table to obtain a first information form.

[0019] Preferably, the gated recurrent unit model based on a deep bidirectional encoder and improved knowledge feedback is combined with a self-attention mechanism to identify and extract quality knowledge from quality documents, including:

[0020] Pre-process the text of engineering quality documents, remove data with missing key information, remove special symbols in the text, and delete invalid words in the dictionary;

[0021] Use a deep bidirectional encoder to fine-tune the text and convert the quality text into a text vector;

[0022] The gated recurrent unit model improved by knowledge feedback extracts features from quality text. The improved model incorporates the construction domain knowledge feature feedback model as follows:

[0023] Δz t =tanh(W k .K q );

[0024] z′ t =λz t +(1-λ)Δz t ;

[0025] Among them, K q is the preset quality standard knowledge vector, λ is the adaptive mixing coefficient, λ=σ(W λ ·[h t-1 ,x t ])), W k is the model parameter of the GRU model, Δz t Indicates the content information of the change, z t is the feature information content at the current moment, z t ' is the feature information content of the next moment;

[0026] Subsequently, an enhanced self-attention mechanism was designed to determine the importance of words in the text, adding orthogonal constraints and joint mapping of quality features:

[0027]

[0028] in, is the feature joint mapping relationship, indicating the importance of words, Q is the quality knowledge prior distribution, Q T K is the standard similarity; α is the learnable coupling parameter, and α||Q T KI|| is an orthogonal constraint, K is the knowledge vector information of the model input; I represents the orthogonal unit matrix;

[0029] Based on the enhanced self-attention mechanism proposed above, technical terms and keywords in quality text are captured to complete the extraction of quality knowledge in quality document text.

[0030] Preferably, the quality knowledge is classified using regular expressions and stored as a quality knowledge table to obtain a second information form, which also includes:

[0031] The extracted quality inspection knowledge is extracted through a regular expression model to obtain the engineering information in the quality inspection knowledge;

[0032] The quality inspection knowledge is classified based on the obtained engineering information, and the classification results are stored as a quality knowledge table to obtain a second information form.

[0033] Preferably, the first and second information forms are merged into a BIM-coupled quality knowledge form through semantic similarity matching of word vectors in three dimensions, thereby obtaining a third information form including:

[0034] Classify the first and second information forms according to the three dimensions of material, project type, and location;

[0035] Calculate the semantic similarity between each dimension in the two forms after classification;

[0036] First, the semantic similarity of the engineering type dimension between the two elements is calculated. If the similarity is equal to 1, the next step is carried out;

[0037] Then calculate the word vector similarity of the position dimension. If the similarity is greater than 0.5, proceed to the next step.

[0038] Finally, calculate the word vector similarity of the material dimension. If the similarity is equal to 1, proceed to the next step;

[0039] Create a mapping relationship between BIM model information and quality inspection knowledge, and output it;

[0040] The output result is stored as a BIM-coupled quality knowledge table to obtain a third information form.

[0041] Preferably, the third information form is constructed into a quality inspection knowledge base using SQL Server to obtain a quality inspection database, which also includes:

[0042] The quality inspection knowledge base contains attribute fields of BIM component information and attribute fields of quality knowledge. The quality knowledge fields and component information fields cannot be empty.

[0043] The knowledge base should also store the mapping relationship between BIM components and quality inspection knowledge, and design query logic, that is, query directly from components to knowledge.

[0044] Preferably, a virtual scene is built in Unity, and a program is written using VS to call the inspection knowledge base in the virtual scene, including:

[0045] In the Unity software development environment, place the BIM model and corresponding markers according to the on-site layout;

[0046] Add different light sources according to scene requirements, adjust the intensity and color of the light sources, and enhance the visual effects of the scene;

[0047] Adjust rendering settings to achieve the best balance between visual quality and performance.

[0048] Preferably, using VS to write a program to call the inspection knowledge base in a virtual scene also includes:

[0049] Use Visual studio 2019 to write scripts and link them to the BIM model to add interactive functions to the model in the scene;

[0050] Create a user interface to provide tools for on-site inquiries for inspectors.

[0051] Preferably, the extracted engineering information is added to the BIM model based on REVIT software, including:

[0052] Save the added BIM model in IFC format;

[0053] Convert IFC format to FBX format and import it into Unity.

[0054] The present invention has the following beneficial effects:

[0055] 1. The present invention provides a method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration. After obtaining the BIM model of the engineering construction, the key information of the BIM model is first extracted to obtain a first information form. Then, NLP technology is used to identify and extract the quality knowledge in the quality file to obtain a second information form. The first and second information forms are then merged through association rules to obtain a third information form. Subsequently, the third information form is constructed into a quality inspection knowledge base using SQL Server to obtain an inspection knowledge database. Finally, a virtual scene is built in Unity, and a program is written using VS to call the inspection knowledge base in the virtual scene. This application utilizes the key information in the constructed BIM model and the quality knowledge extracted by NLP technology, and couples the BIM model and quality inspection knowledge based on the semantic similarity between the three corresponding dimensions, thereby reducing the time for manual query and search and improving the efficiency of quality inspection.

[0056] 2. The present invention associates the BIM model with quality inspection knowledge, and can query inspection knowledge through the model in the augmented reality scene, which simplifies the steps of inspection knowledge query and improves the accuracy of quality inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The present invention will be further described below with reference to the accompanying drawings and examples.

[0058] Figure 1 A flowchart of a method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration is provided in an embodiment of the present application.

[0059] Figure 2A flowchart for extracting key information of BIM model components provided in an embodiment of the present application.

[0060] Figure 3 A flowchart of extracting inspection knowledge from quality documents using natural language processing technology provided in an embodiment of the present application.

[0061] Figure 4 A schematic diagram of a construction quality inspection database provided in an embodiment of the present application.

[0062] Figure 5 A construction BIM model is provided in an embodiment of the present application.

[0063] Figure 6 Schematic diagram of the process of adding parameters to a BIM model.

[0064] Figure 7 A schematic diagram of the association rules and judgment method for the integration of construction quality inspection knowledge and BIM provided in an embodiment of the present application. DETAILED DESCRIPTION

[0065] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0066] Example 1:

[0067] First, combine Figure 1 The present invention provides a method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration, as shown in the following example. Figure 1 As shown, the method may include:

[0068] Step S01: Add engineering information to the BIM model based on REVIT software.

[0069] Specifically, collect the design drawings of the on-site construction buildings, build a BIM model of the construction site through Revit software, and use this as the construction BIM model for other steps. For example, take a sluice pump under construction in a certain project as an example, build a BIM model for it, and get the following Figure 5 The construction BIM model shown.

[0070] Step S02: Based on the IfcOpenShell library, define component entity fields, automatically extract BIM key information, and store it as a BIM information table.

[0071] Specifically, if Figure 2As shown, based on the acquired construction BIM model, the added component engineering information is determined, the parameters are added to the BIM model based on the Revit software, and the model format is converted to IFC to serve as the initial BIM model. The initial BIM model includes detailed dimensions, materials, names, project types, locations and other information of the project. Then, the extracted information and parameters are defined, and a python program is written based on the IfcOpenShell library to extract the key information stored in the initial BIM model in the IFC format, store it as a BIM model information form, and obtain the first information form. For example: Build a BIM model of the laboratory building according to the gate pump design file, and add information parameters such as Figure 6 shown.

[0072] Step S03: Based on the deep bidirectional encoder and the gated recurrent unit model improved by knowledge feedback, the self-attention mechanism is combined to identify and extract the quality knowledge in the quality file and store it as an inspection knowledge table.

[0073] Specifically, if Figure 3 As shown in the figure, the engineering quality document text is preprocessed, which includes three steps: eliminating data with missing key information, removing special symbols in the text, and deleting invalid words in the dictionary; then using the deep bidirectional encoder (BERT) to fine-tune the text and convert the quality text into a text vector; and using the gated recurrent unit model improved by knowledge feedback to screen the valid content in the text and extract features from the quality text.

[0074] The GRU model introduces two gate mechanisms—update gate and reset gate to determine which feature information should be passed and output. t Decide how much information to keep from the past hidden state; reset gate r t Determines how much of the current hidden state needs to be eliminated when calculating the current hidden state. The input of the GRU model is the quantized text result W t and hidden state h t-1 Therefore, the update gate and reset gate are calculated as follows:

[0075]

[0076] Among them, G xr , G hr , G xz is the weight matrix, σ() is the sigmoid function.

[0077] The improved model incorporates the construction domain knowledge feature feedback model as follows:

[0078] Δz t =tanh(W k .K q );

[0079] z′ t =λz t +(1-λ)Δz t ;

[0080] Among them, K q is the preset quality standard knowledge vector, λ is the adaptive mixing coefficient (λ=σ(W λ ·[h t-1 ,x t ]));W k is the model parameter of the GRU model, Δz t Indicates the content information of the change, z t is the feature information content at the current moment, z t ' is the feature information content of the next moment;

[0081] Use the update gate information and candidate hidden state to update the content of GRU and get the final hidden state h t as follows:

[0082]

[0083] In the formula, (1-z t )·h t-1 Valuable information saved for the GRU; h is the selection information of candidate hidden states; t is the final hidden state, indicating how much existing information can be retained for the next GRU.

[0084] Subsequently, an enhanced self-attention mechanism was designed to determine the importance of words in the text, adding orthogonal constraints and joint mapping of quality features:

[0085]

[0086] in, is the feature joint mapping relationship, indicating the importance of words, Q is the quality knowledge prior distribution, Q T K is the standard similarity; α is the learnable coupling parameter, and α||Q T KI|| is an orthogonal constraint, K is the knowledge vector information of the model input; I represents the orthogonal unit matrix.

[0087] The enhanced self-attention mechanism proposed above captures technical terms and keywords in quality documents, extracting important quality inspection knowledge from the text. This method then preserves this knowledge within the quality documents. A regular expression (RE) model is used to extract this knowledge, revealing the engineering information within it. This information is then categorized into three dimensions. The resulting quality inspection knowledge is then classified based on these three dimensions and stored as a quality knowledge table, serving as the second information table.

[0088] Step S04: BIM information and inspection knowledge are merged into a BIM-coupled quality knowledge table through word vector semantic similarity matching in three dimensions.

[0089] Specifically, according to the association rules, the first information table extracted from the BIM model and the quality knowledge table extracted from the quality document text are linked, and feature matching is performed according to the three dimensions respectively.

[0090] In order to perform feature matching, the similarity between semantic feature information is calculated to achieve fuzzy matching of BIM component quality knowledge. The feature matching process uses text word vectors to calculate the semantic similarity between words and determine the degree of relevance, which is defined as:

[0091]

[0092] Where W K Represents the word vector of the kth keyword retrieved from the quality knowledge, W j It represents the word vector of the j-th keyword of the same dimension extracted from the BIM component information, and S(kj) represents the semantic similarity between word k and word j.

[0093] Based on this semantic similarity calculation, an association rule for matching quality knowledge with BIM model information is proposed. The judgment method is as follows: Figure 7 As shown, the main steps are as follows:

[0094] (1) Input the feature word vector W extracted from the BIM model information j .

[0095] (2) Input the feature word vector W in quality knowledge k .

[0096] (3) Calculate the similarity of the keyword vectors in the engineering type dimension. If the semantic similarity is 1, proceed to the next step; otherwise, return to step (1).

[0097] (4) Calculate the keyword vector similarity in the position dimension. If the word vector semantic similarity is greater than or equal to 0.5, continue to step (6), otherwise proceed to step (5).

[0098] (5) Calculate the semantic similarity of the keyword vectors in the material dimension. If the word vector similarity is 1, continue to step (6), otherwise return to step 1○.

[0099] (6) Create links between BIM components and quality knowledge and combine them to output a third information form.

[0100] Step S05: Build a quality inspection knowledge base using SQL Server.

[0101] Specifically, based on SQL Server, the third information form is constructed into a quality inspection database as a quality inspection knowledge base. The quality inspection knowledge base contains attribute fields of BIM component information and attribute fields of quality knowledge, wherein the quality knowledge field and component ID field cannot be empty.

[0102] Step S06: Build a virtual-reality scene in Unity, and use VS to write a program to call the knowledge base in the virtual scene.

[0103] Specifically, to run in an augmented reality scenario, it is first necessary to build a virtual and real scene in Unity's virtual environment based on the on-site location. Real-time data retrieval and knowledge base updates are performed in the virtual scene. Using a Web API as an intermediate component, the virtual and real scene can securely access and manipulate data in the knowledge base. This is accomplished using a three-tier architecture:

[0104] (1) Quality inspection knowledge database: This is the knowledge base constructed above, which is used to store BIM information and quality knowledge. It is also the data storage layer and is hosted on a cloud server or a physical IoT server to achieve network operations.

[0105] (2) Web API: As an intermediate component, Web API is responsible for directly interacting with the knowledge base and calling endpoints through the HTTP network to achieve global access to the database of virtual scenes.

[0106] (3) Augmented reality inspection application: Through the construction and deployment of Unity, a virtual and real augmented inspection application is realized, and data is retrieved from the database to dynamically update the virtual and real scenes.

[0107] Based on the above operations, the construction of the augmented reality construction quality inspection knowledge base is completed.

[0108] The method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration provided by this application is to extract the key information of the BIM model after obtaining the BIM model of the engineering construction to obtain the first information form. Then, NLP technology identifies and extracts the quality knowledge in the quality file to obtain the second information form. The first and second information forms are then merged through association rules to obtain the third information form. Subsequently, the third information form is constructed into a quality inspection knowledge base using SQL Server to obtain an inspection knowledge database. Finally, a virtual scene is built in Unity, and a program is written using VS to call the inspection knowledge base in the virtual scene. This application utilizes the key information in the constructed BIM model and the quality knowledge extracted by NLP technology, and couples the BIM model and quality inspection knowledge based on the semantic similarity between the three corresponding dimensions, thereby reducing the time for manual query and search and improving the efficiency of quality inspection.

[0109] This application associates BIM models with quality inspection knowledge, and can query inspection knowledge through models in virtual-reality fusion scenes, which simplifies the steps of inspection knowledge query and improves the accuracy of quality inspection.

Claims

1. A method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration, characterized in that: The method comprises: Extract model-related engineering information from construction design drawings and design documents, and add the extracted engineering information to the BIM model using REVIT software; Based on the IfcOpenShell library, component entity fields are defined, BIM key semantics are automatically extracted, and stored as a BIM information table to obtain the first information form; Based on a deep bidirectional encoder and a gated recurrent unit model improved by knowledge feedback, the self-attention mechanism is combined to identify and extract quality knowledge from quality documents. Regular expressions are used to classify the quality knowledge and store it in a quality knowledge table to obtain the second information form. By matching the semantic similarity of word vectors in three dimensions, the first and second information forms are merged into a BIM-coupled quality knowledge table to obtain the third information form; Using SQL Server to construct the third information form into a quality inspection knowledge base, thus obtaining a quality inspection database; Build a virtual and real scene in Unity, and use VS to write a program to call the inspection knowledge base in the virtual scene.

2. The method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration according to claim 1 is characterized in that: Extract model-related engineering information from construction design drawings and design documents, and add the extracted engineering information to the BIM model using REVIT software, including: Leveraging computer vision and deep learning technologies, we automatically extract key engineering information from engineering drawings, including the type of project, physical locations, and the relationships and connections between elements. Open the building model to which you need to add parameters, select the corresponding component, add the project parameters, and associate the parameters with the element through the SetParameter method of the Element class.

3. The method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration according to claim 1 is characterized in that: Based on the IfcOpenShell library, component entity fields are defined, BIM key semantics are automatically extracted, and stored as a BIM information table to obtain the first information form, including: Define key information parameters in the BIM model and convert the BIM model into IFC format file; Based on the IfcOpenShell library, a Python program is used to filter the model component entities in the IFC file, read the attribute fields of these component entities, and obtain the key information of each model component; The filtered and extracted component key information is sorted and stored as a BIM information table to obtain a first information form.

4. The method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration according to claim 1 is characterized in that: Based on a deep bidirectional encoder and a gated recurrent unit model improved by knowledge feedback, the self-attention mechanism is combined to identify and extract quality knowledge from quality documents, including: Pre-process the text of engineering quality documents, remove data with missing key information, remove special symbols in the text, and delete invalid words in the dictionary; Use a deep bidirectional encoder to fine-tune the text and convert the quality text into a text vector; The gated recurrent unit model improved by knowledge feedback extracts features from quality text. The improved model incorporates the construction domain knowledge feature feedback model as follows: Δz t =tanh(W k .K q ); With t ' =λz t +(1-λ)Δz t ; Among them, K q is the preset quality standard knowledge vector, λ is the adaptive mixing coefficient, λ=σ(W λ ·[h t-1 ,x t ])), W k is the model parameter of the GRU model, Δz t Indicates the content information of the change, z t is the feature information content at the current moment, z′ t Characteristic information content for the next moment; Subsequently, an enhanced self-attention mechanism was designed to determine the importance of words in the text, adding orthogonal constraints and joint mapping of quality features: in, is the feature joint mapping relationship, indicating the importance of words, Q is the quality knowledge prior distribution, Q T K is the standard similarity; α is the learnable coupling parameter, and α||Q T KI|| is an orthogonal constraint, K is the knowledge vector information of the model input; I represents the orthogonal unit matrix; Based on the enhanced self-attention mechanism proposed above, technical terms and keywords in quality text are captured to complete the extraction of quality knowledge in quality document text.

5. The method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration according to claim 4 is further characterized in that: The quality knowledge is classified using regular expressions and stored as a quality knowledge table to obtain a second information form, which also includes: The extracted quality inspection knowledge is extracted through a regular expression model to obtain the engineering information in the quality inspection knowledge; The quality inspection knowledge is classified based on the obtained engineering information, and the classification results are stored as a quality knowledge table to obtain a second information form.

6. The method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration according to claim 1 is characterized in that: Through semantic similarity matching of word vectors in three dimensions, the first and second information forms are merged into a BIM-coupled quality knowledge table, resulting in a third information form, including: Classify the first and second information forms according to the three dimensions of material, project type, and location; Calculate the semantic similarity between each dimension in the two forms after classification; First, the semantic similarity of the engineering type dimension between the two elements is calculated. If the similarity is equal to 1, the next step is carried out; Then calculate the word vector similarity of the position dimension. If the similarity is greater than 0.5, proceed to the next step. Finally, calculate the word vector similarity of the material dimension. If the similarity is equal to 1, proceed to the next step; Create a mapping relationship between BIM model information and quality inspection knowledge, and output it; The output result is stored as a BIM-coupled quality knowledge table to obtain a third information form.

7. The method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration according to claim 1 is characterized in that: The third information form is constructed into a quality inspection knowledge base using SQL Server to obtain a quality inspection database, which also includes: The quality inspection knowledge base contains attribute fields of BIM component information and attribute fields of quality knowledge. The quality knowledge fields and component information fields cannot be empty. The knowledge base should also store the mapping relationship between BIM components and quality inspection knowledge, and design query logic, that is, query directly from components to knowledge.

8. The method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration according to claim 1 is characterized in that: Build a virtual scene in Unity and use VS to write a program to call the inspection knowledge base in the virtual scene, including: In the Unity software development environment, place the BIM model and corresponding markers according to the on-site layout; Add different light sources according to scene requirements, adjust the intensity and color of the light sources, and enhance the visual effects of the scene; Adjust rendering settings to achieve the best balance between visual quality and performance.

9. The method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration according to claim 8 is further characterized in that: Using VS to write a program to call the inspection knowledge base in a virtual scene also includes: Use Visual studio 2019 to write scripts and link them to the BIM model to add interactive functions to the model in the scene; Create a user interface to provide tools for on-site inquiries for inspectors.

10. The method for constructing a construction quality inspection knowledge base based on natural language extraction and BIM integration according to claim 1, characterized in that: The extracted engineering information is added to the BIM model based on REVIT software, including: Save the added BIM model in IFC format; Convert IFC format to FBX format and import it into Unity.

Images