A BIM model information auditing method, device and computer program

Abstract

The application provides a BIM model information auditing method, device and computer program, for the data-cleaning building information model, the pre-trained semantic recognition model is used to extract the semantics of the name, function description and specification description of each data field in the building information model to be audited, the semantics of the data field name is used for comparison with the semantics of the standard field name collected in the standard attribute field library, and the data field contained in the standard attribute field library in the building information model to be audited is extracted to construct a building information model attribute value library to be audited, so that the data field is accurately aligned at the semantic level; further based on the constraints recorded in the attribute value rule library, the values or description texts recorded in each data field in the building information model attribute value library to be audited are checked, the auditing items are enriched, and the constraint mode of the values or texts in the data field in the auditing process is deepened.

Description

Technical Field

[0001] This invention relates to the field of information technology in construction engineering, and in particular to a method, apparatus and computer program for reviewing BIM model information. Background Technology

[0002] Building Information Modeling (BIM) is a digital model based on a three-dimensional model that contains all relevant information about a building. This model not only helps designers better understand and design buildings, but can also be used in the construction, operation, and maintenance phases.

[0003] Currently, model information review mainly relies on manual inspection, which suffers from high workload and insufficient accuracy. The goal of automated BIM model review software is to improve work efficiency and quality by automatically and quickly reviewing BIM models through computers to meet standard and specification requirements. Current BIM model review software primarily performs information integrity checks. However, faced with increasingly complex BIM model information, existing review solutions can only perform integrity verification, lacking sufficient depth of review. Furthermore, it is insufficient to achieve standardized review of information values ​​in the face of diverse field annotations and textual descriptions. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a BIM model information review method, apparatus and computer program to eliminate or improve one or more defects existing in the prior art, and solve the problem in the prior art that the inconsistency in the text expression in the BIM model leads to the inability to align and perform in-depth verification.

[0005] One aspect of the present invention provides a BIM model information verification method, the method comprising the following steps:

[0006] Obtain the building information model to be reviewed, which includes geometric and non-geometric information recorded based on data fields; the geometric information includes the location, shape, and size information of the building object; the non-geometric information includes the material, specifications, cost, and construction time information of the building object;

[0007] The building information model to be reviewed is subjected to data preprocessing, which includes data cleaning and format standardization.

[0008] Obtain a pre-trained semantic recognition model, and extract the semantics of the data field names in the pre-processed building information model to be reviewed to obtain the corresponding semantics of the field names to be reviewed.

[0009] Obtain a standard attribute field library for building information model, compare the semantics of the field name to be reviewed with the semantics of the standard field name of each data field in the standard attribute field library, extract the data fields contained in the standard attribute field library of the building information model to be reviewed, and construct an attribute value library of the model to be reviewed. The structure of the attribute value library of the model to be reviewed is the same as that of the standard attribute field library.

[0010] Obtain the attribute value rule library related to the standard attribute field library, and review the values ​​of each data field in the attribute value library of the model to be reviewed based on the intra-field constraint rules and inter-field relationship constraint rules recorded in the attribute value rule library to obtain the target review result;

[0011] A summary report on whether the target audit results meet the standards is generated and sent to the target audit port via a preset link.

[0012] In some embodiments, the non-geometric information further includes textual description information, which includes a functional description and specification description of the building;

[0013] Semantic extraction is performed on the functional description to obtain the semantics of the function to be reviewed, and semantic extraction is performed on the specification description to obtain the semantics of the specification to be reviewed;

[0014] The consistency of the functional semantics corresponding to the textual description information in the attribute value library of the model to be audited is compared with the standard functional semantics recorded in the standard attribute field library. The consistency of the normative semantics corresponding to the textual description information in the attribute value library of the model to be audited is compared with the standard normative semantics recorded in the standard attribute field library. The target audit result is then expanded based on the comparison results.

[0015] In some embodiments, the pre-training step of the semantic recognition model includes:

[0016] Obtain a training sample set, which includes multiple positive and negative samples. The positive samples contain semantically consistent text pairs within the domain related to the building information model, and the negative samples contain semantically inconsistent text pairs within the domain related to the building information model. Add a label to the positive and negative samples indicating whether they are semantically consistent.

[0017] An initial neural network model is obtained, which employs a skip-garm model and a similarity calculation module. The skip-garm model performs semantic extraction on text pairs in the positive or negative samples, and the similarity calculation module calculates the similarity of text pairs to obtain the prediction results of consistency comparison.

[0018] The initial neural network model is trained using the training sample set. The cross-entropy loss function is calculated based on the prediction results and the labels. The parameters of the initial neural network model are updated based on the cross-entropy loss function. The updated skip-garm model is then used as the semantic recognition model.

[0019] In some embodiments, the in-field constraint rules include:

[0020] Constraint rules within string fields are used to check the string's length, format, value range, uniqueness, whether it is required, and whether there are prohibited characters.

[0021] Numerical field constraint rules are used to check the range, precision, scale, and unit of the numerical values ​​in the data field.

[0022] Constraint rules within enumerated fields are used to check whether the enumerated values ​​within the data field are included within a preset range;

[0023] Boolean field constraint rules are used to check the value form, non-nullability, and dependency on the target field of the Boolean values ​​in the data field.

[0024] Constraint rules within the date field are used to verify the format and sequence of date data within the data field.

[0025] In some embodiments, the method further includes: calculating the entity length and entity position of the building information model to be reviewed based on the geometric information, comparing it with the attribute length and attribute position of a standard building information model, generating a location review result, and sending it to the target review port according to the preset link.

[0026] In some embodiments, the method further includes: when the target audit result contains problematic data fields that do not conform to the field constraint rules and the inter-field relationship constraint rules, searching for other data field contents that generated the problematic data fields, and regenerating and updating the problematic data fields.

[0027] In some embodiments, the method is further used to calculate an overall accuracy rate as a review result, including:

[0028] Each item in the target review results is assigned a weight based on review difficulty and workload, calculated as follows:

[0029] ;

[0030] in, This represents the weight of the i-th entry. This represents the construction difficulty of the i-th entry. This represents the workload of the i-th entry; n represents the number of model files;

[0031] The precision rate for the target project is calculated using the following formula:

[0032] ;

[0033] Where P is the precision rate of the target project. This represents the precision of the i-th entry in the j-th model file; m represents the total number of entries in the j-th model file.

[0034] The recall rate for the target project is calculated using the following formula:

[0035]

[0036] Where R is the recall rate of the target project. This represents the recall of the i-th entry in the j-th model file; m represents the total number of entries in the j-th model file.

[0037] Calculate the overall accuracy of the target project. The calculation formula is:

[0038] .

[0039] On the other hand, the present invention also provides a BIM model information review device, including a processor, a memory, and a computer program / instructions stored in the memory. The processor is used to execute the computer program / instructions, and when the computer program / instructions are executed, the device implements the steps of the above method.

[0040] On the other hand, the present invention also provides a computer-readable storage medium having a computer program / instructions stored thereon, characterized in that the computer program / instructions, when executed by a processor, implement the steps of the above-described method.

[0041] On the other hand, the present invention also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described method.

[0042] The beneficial effects of the present invention are at least as follows:

[0043] The BIM model information review method, device, and computer program of this invention, for the BIM model to be reviewed after data cleaning, utilize a pre-trained semantic recognition model to semantically extract the names, functional descriptions, and specification descriptions of each data field in the BIM model to be reviewed. The semantics of the data field names are compared with the semantics of the standard field names summarized in the standard attribute field library, and the data fields contained in the standard attribute field library of the BIM model to be reviewed are extracted to construct the attribute value library of the model to be reviewed, thus achieving accurate alignment of data fields at the semantic layer. Furthermore, based on the constraints recorded in the attribute value rule library, the numerical values ​​or descriptive text recorded in each data field of the attribute value library of the model to be reviewed are checked, enriching the review items and deepening the constraint methods on the numerical values ​​or text in the data fields during the review process.

[0044] Furthermore, by configuring corresponding constraint rules for different forms of numerical values ​​within the data fields, we can adapt to the diverse data formats required for BIM model review, thereby improving efficiency and review accuracy.

[0045] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the description, or may be learned by practice of the invention. The objects and other advantages of the invention can be realized and obtained by means of the structures specifically pointed out in the description and drawings.

[0046] Those skilled in the art will understand that the objectives and advantages achievable with the present invention are not limited to those specifically described above, and that the above and other objectives achievable with the present invention will become clearer from the following detailed description. Attached Figure Description

[0047] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, are not intended to limit the scope of the invention. In the drawings:

[0048] Figure 1 This is a flowchart illustrating the BIM model information review method according to an embodiment of the present invention.

[0049] Figure 2 This is a schematic diagram illustrating the logical relationship of the BIM model information review method according to an embodiment of the present invention.

[0050] Figure 3 This is a schematic diagram illustrating the review process of the BIM model information review method according to an embodiment of the present invention.

[0051] Figure 4 This is a software architecture diagram of the BIM model automatic review tool according to another embodiment of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0053] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.

[0054] It should be emphasized that the term "including / comprises" as used herein refers to the presence of a feature, element, step, or component, but does not exclude the presence or addition of one or more other features, elements, steps, or components.

[0055] Automated review of BIM models is an urgent need in this field. Current BIM model review software mainly focuses on information integrity. This involves building a database of standard requirements to identify all information fields required for various types of components. Then, BIM information is extracted to create a review model database, compiling the information fields already added to various models. By comparing and verifying these two databases, missing or incorrect attribute information in the BIM model is reported. However, given the diverse forms and logical relationships of current building information models, the difficulty in review lies in the fact that during the compilation, use, and recording processes by multiple stakeholders, the names and descriptions of certain items are often colloquial and non-standardized, and the order and structure of data fields may also differ. This leads to a reliance on manual review, and even with some degree of automation, it is difficult to address the challenges of information extraction due to non-standardized data field structures, diverse comparison rules, and difficulties in review due to non-standardized text records.

[0056] This invention provides a method for reviewing BIM model information, such as... Figure 1 As shown, the method includes the following steps S101~S106:

[0057] Step S101: Obtain the building information model to be reviewed. The building information model to be reviewed includes geometric and non-geometric information recorded based on data fields. The geometric information includes the location, shape, and size information of the building object. The non-geometric information includes the material, specifications, cost, and construction time information of the building object.

[0058] Step S102: Perform data preprocessing on the building information model to be reviewed. Data preprocessing includes data cleaning and format standardization.

[0059] Step S103: Obtain the pre-trained semantic recognition model, and extract the semantics of the data field names in the pre-processed building information model to be reviewed to obtain the corresponding semantics of the field names to be reviewed.

[0060] Step S104: Refer to Figure 2 Obtain the standard attribute field library for the building information model, compare the semantics of the field names to be reviewed with the semantics of the standard field names of each data field in the standard attribute field library, and extract the data fields contained in the standard attribute field library of the building information model to be reviewed, and construct the attribute value library of the model to be reviewed. The structure of the attribute value library of the model to be reviewed is the same as that of the standard attribute field library.

[0061] Step S105: Refer to Figure 2 and Figure 3 The system retrieves the attribute value rule library related to the standard attribute field library. Based on the intra-field constraint rules and inter-field relationship constraint rules recorded in the attribute value rule library, the system audits the values ​​of each data field in the attribute value library of the model to be audited, and obtains the target audit result.

[0062] Step S106: Generate a summary report on whether the target audit results meet the standards, and send it to the target audit port according to the preset link.

[0063] In step S101, the geometric information includes: position (the building's specific coordinates in space, typically represented by XYZ coordinates); shape (the building's three-dimensional form, including the specific shapes and structures of walls, roofs, windows, doors, etc.); and dimensions (the building's or its components' length, width, height, and other specific dimensions).

[0064] In the non-geometric information, "materials" represents the type and properties of materials used in the building, such as reinforced concrete, glass, and wood. "Specifications" represents the specific dimensions of the materials, such as the diameter of the reinforcing bars and the thickness of the glass. "Costs" represents various cost information related to the building, including material costs and labor costs. "Construction time" refers to the time schedule for the building's construction as recorded in the model, including the start time, end time, and the progress plan for each stage.

[0065] In step S102, in some embodiments, the data preprocessing includes filling missing data with specified values ​​and performing unit conversion on the values ​​recorded in each data field.

[0066] Specifically, the purpose of data cleaning is to remove or repair erroneous, redundant, or inconsistent data in the BIM model to improve data quality and usability. This includes the following aspects: 1) Data integrity check: Checking for missing attributes or information in the model, such as missing geometric information, material properties, or component information. Missing data is completed or marked. 2) Data consistency check: Ensuring that similar data uses consistent naming conventions and units, such as using the same naming rules and unit system (meters, millimeters, etc.) for the same type of components. 3) Duplicate data processing: Identifying and deleting duplicate components or redundant data in the model to reduce data redundancy. 4) Error data correction: Finding and correcting errors in the data, such as geometric errors (e.g., face intersections, non-closed surfaces), attribute errors (e.g., incorrect material properties), etc. 5) Data filtering: Filtering out data that needs to be reviewed based on audit requirements, and removing irrelevant or unnecessary data.

[0067] The purpose of format standardization is to ensure that model data conforms to certain standards and specifications, enabling its correct transfer and use across different software and platforms. Specifically, this includes: 1) Unit and precision standardization: converting all data in the model to a unified audit system and precision requirements. 2) Attribute name standardization: unifying component attribute names according to industry or project standards, such as using standardized material names and classification codes. 3) File format conversion: converting model data to the file format required for auditing, ensuring consistency comparison during the audit process. 4) Geometric precision adjustment: adjusting the geometric precision of the model according to standards or audit requirements, such as adjusting the level of modeling details, simplifying or refining geometric structures. 5) Coordinate system and reference point standardization: unifying the coordinate system and reference point settings of the model to ensure correct positioning and alignment in different environments.

[0068] In step S103, in some embodiments, the pre-training step of the semantic recognition model includes S1031~S1033:

[0069] Step S1031: Obtain the training sample set, which includes multiple positive and negative samples. Positive samples contain semantically consistent text pairs within the domain related to Building Information Modeling (BIM), while negative samples contain semantically inconsistent text pairs within the domain related to BIM. Add labels to the positive and negative samples indicating whether they are semantically consistent.

[0070] Step S1032: Obtain the initial neural network model. The initial neural network model uses the skip-garm model and a similarity calculation module. The skip-garm model performs semantic extraction on text pairs in positive or negative samples. The similarity calculation module calculates the similarity of text pairs and obtains the prediction results of consistency comparison.

[0071] Step S1033: Train the initial neural network model using the training sample set, calculate the cross-entropy loss function based on the prediction results and the labels, update the parameters of the initial neural network model based on the cross-entropy loss function, and use the updated skip-garm model as the semantic recognition model.

[0072] Steps S1031-S1033 of this application train a semantic recognition model for technical terms in the field of Building Information Modeling (BIM). The Skip-Gram model used in this process is a neural network model for generating word embeddings, a method for converting words into continuous vector representations. These vectors reside in a high-dimensional space and, to some extent, preserve the semantic relationships between words. Here, the Skip-Gram model is used to extract semantic features from text pairs in each positive or negative sample. The similarity calculation module receives the semantic expressions obtained from the word embedding operation of the Skip-Gram model and calculates the similarity between text pairs. Based on this similarity, the model can generate a prediction result, determining whether the text pairs are semantically consistent or inconsistent. During the recognition of positive and negative samples, a cross-entropy loss is constructed to update the model parameters. The updated Skip-Gram model then serves as the semantic recognition model, performing word embedding for semantic extraction and comparison.

[0073] In step S104, based on semantic extraction, the semantics of the field names to be reviewed are compared with the semantics of the standard field names in the standard attribute field library. Data fields contained in the standard attribute field library within the building information model to be reviewed are extracted and used to construct an attribute value library for the model to be reviewed. This is primarily to align the data fields and maintain consistency with the structure of the standard attribute field library, facilitating subsequent review. In step S105, the attribute value rule library is used to record various constraint rules for the standard attribute field library, so that review can be conducted according to the corresponding constraint rules. Specifically, for numerical data fields, the in-field constraint rules are configured as follows:

[0074] Constraint rules within string fields are used to check the string's length, format, value range, uniqueness, whether it is required, and whether there are prohibited characters.

[0075] Numeric field constraint rules are used to check the range, precision, scale, and unit of values ​​within a data field.

[0076] Constraint rules within enumerated fields are used to check whether enumerated values ​​within a data field are included within a preset range.

[0077] Boolean field constraint rules are used to check the value form, non-nullability, and dependency on target fields of Boolean values ​​within data fields.

[0078] Constraint rules within date fields are used to verify the format and chronological order of date data within the data field.

[0079] The rules governing relationships between fields mainly include constraints formed between different data fields based on logical relationships, operational relationships, and normative comparison relationships.

[0080] Logical relationship constraints refer to the relationships that must be followed between different data fields based on business logic or practical application scenarios. These constraints ensure that the data between fields conforms to specific business logic. Examples include: conditional dependency, where the value of one field determines the possible range or value of another field; and mutual exclusion, where certain fields are mutually exclusive, meaning that if one field has a value, the other field must be empty.

[0081] Operational constraints refer to the requirement that different data fields must satisfy a certain mathematical or arithmetic relationship. These constraints are typically applied to numeric fields to ensure that the operational relationships between the data are reasonable. For example, the value of the "Total Amount" field should equal the result of "Unit Price" multiplied by "Quantity". Or, the value of one field must be within a range of another field's values; for example, the value of the "Age" field must be greater than or equal to the difference between "Year of Birth" and the current year.

[0082] Normative comparison constraints refer to the comparison relationships between different data fields based on industry standards, regulations, or internal specifications. These constraints ensure that data conforms to specific standards or specifications. For example, in a code lookup table, the values ​​of certain fields need to be compared and verified against a standardized code table or reference table. Format consistency requires field values ​​to conform to specific formats or specifications, such as the format of postal codes or telephone numbers. If a field value conforms to a specific format, then related fields must also be adjusted accordingly. For example, if an "ID number" field is an 18-digit number, then the "date of birth" field should conform to the year information contained in the ID number.

[0083] Furthermore, for textual descriptive information, similarity comparison can be directly performed by extracting semantics. Specifically, non-geometric information also includes textual descriptive information, which includes functional descriptions and regulatory descriptions of the building. Functional descriptions are detailed descriptions of the building's function, such as whether a building is used as an office building, residence, or shopping mall. Regulatory descriptions are descriptions of technical specifications related to the building's design and construction, such as seismic design codes and energy-saving standards.

[0084] Semantic extraction is performed on the functional description to obtain the semantics of the function to be reviewed, and semantic extraction is performed on the specification description to obtain the semantics of the specification to be reviewed;

[0085] The consistency of the semantics of the function to be audited corresponding to the text description information in the attribute value library of the model to be audited is compared with the standard semantics of the function recorded in the standard attribute field library. The consistency of the semantics of the specification to be audited corresponding to the text description information in the attribute value library of the model to be audited is compared with the standard specification semantics recorded in the standard attribute field library. The target audit result is then expanded based on the comparison results.

[0086] In step S106, based on the review of all constraints recorded in the attribute value rule base in step S105, the review result of each review item is obtained, i.e., whether it meets or does not meet the requirements. Further statistical analysis can be performed to obtain the corresponding review result.

[0087] To integrate the review results of each item into the final review result, the following strategies can be adopted to adapt to different review scenarios and levels of rigor:

[0088] The single-item veto strategy is relatively strict and suitable for high-requirement scenarios. If any audit item results in "non-compliance," the final audit result is "fail." It is suitable for critical areas such as security audits, legal compliance audits, and quality management. The strategy rule is that if any item's audit result is "non-compliance," the final result is "fail."

[0089] The percentage-based pass / fail strategy allows for some flexibility and is suitable for more lenient review scenarios. If the percentage of approved items reaches a preset threshold, the final review result is "passed"; otherwise, it is "failed." This strategy is suitable for non-critical business reviews, comprehensive evaluation systems, etc.

[0090] The weighted scoring strategy assigns different weights to each review item, calculates a weighted total score, and compares the total score with a set threshold to arrive at the final review result. This strategy is suitable for situations where different items have varying degrees of importance. It is applicable to scenarios such as multi-dimensional comprehensive evaluation, performance appraisal, and complex system assessment. The strategy rules include: assigning a weight (e.g., 0.1, 0.2, 0.3, etc.) to each item based on its importance; calculating the weighted total score for all items (items that meet the requirements receive full marks, and those that do not receive 0 points); setting a passing threshold (e.g., a total score of 70%); if the weighted total score is greater than or equal to the threshold, the final result is "pass"; otherwise, it is "fail".

[0091] A classification and tiering strategy categorizes items into different levels or tiers based on their importance (e.g., "critical," "important," "general"), with different passing criteria for each level. This strategy is suitable for multi-level risk control and is applicable to scenarios such as risk management of complex systems and project acceptance.

[0092] Customizable combination strategies allow you to combine some or all of the above strategies based on the needs of specific scenarios. For example, a "single-item veto strategy" can be used for critical items, while a "percentage-based pass strategy" can be used for general items, combining both to arrive at the final review result. This approach is suitable for customized review scenarios and review systems that consider multiple conditions comprehensively.

[0093] In some embodiments, the method further includes step S107: calculating the entity length and entity offset of the building information model to be reviewed based on geometric information, comparing it with the attribute length and attribute position of the standard building information model, generating a location review result, and sending it to the target review port according to a preset link.

[0094] In some embodiments, the method further includes step S108: when the target audit result has a problematic data field that does not conform to the field constraint rules and the inter-field relationship constraint rules, find the contents of other data fields that generated the problematic data field, and regenerate and update the problematic data field.

[0095] In some embodiments, the method is further used to decompose the target engineering project into multiple model files, each model file containing multiple sub-building information models, and to calculate the overall accuracy rate as the review result, including steps S201~S204:

[0096] Step S201: Assign weights to each item in the target audit results according to audit difficulty and workload. The calculation formula is as follows:

[0097] ;

[0098] in, This represents the weight of the i-th entry. This represents the building difficulty of the i-th entry. This represents the workload of the i-th entry; n represents the number of model files.

[0099] Step S202: Calculate the precision rate for the target project. The calculation formula is:

[0100] ;

[0101] Where P represents the precision rate of the target project. represents the precision of the i-th entry in the j-th model file; m represents the total number of entries in the j-th model file.

[0102] Step S203: Calculate the recall rate for the target project. The calculation formula is:

[0103]

[0104] Where R is the recall rate of the target project. This represents the recall of the i-th entry in the j-th model file; m represents the total number of entries in the j-th model file.

[0105] Step S204: Calculate the overall accuracy of the target project. The calculation formula is:

[0106] .

[0107] On the other hand, the present invention also provides a BIM model information review device, including a processor, a memory, and a computer program / instructions stored in the memory. The processor is used to execute the computer program / instructions, and when the computer program / instructions are executed, the device implements the steps of the above method.

[0108] On the other hand, the present invention also provides a computer-readable storage medium having a computer program / instructions stored thereon, characterized in that the computer program / instructions, when executed by a processor, implement the steps of the above-described method.

[0109] On the other hand, the present invention also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described method.

[0110] The present invention will now be described with reference to a specific embodiment:

[0111] This embodiment provides a BIM model information review method, aiming to strengthen the review of model information standardization and adding a reminder function for non-standard BIM model attribute values ​​and logical errors. Compared with traditional BIM model review tools, it adds a model attribute value library and an attribute value rule library. After completing the attribute field check, it can check the attribute values ​​of correct attribute fields according to the attribute value rule library.

[0112] Reference Figure 2 The BIM model automatic review tool comprises four modules: a standard attribute field library, a model attribute field library, an attribute value library, and an attribute value rule library. These modules operate sequentially. The review software first reads the standard file and creates the standard attribute field library. Then, it reads the building information model file to be reviewed, creates the model attribute field library, and compares the model attributes with the standard fields. Next, it extracts the correctly identified model attributes and saves them as the attribute value library. Finally, based on the pre-created attribute value rule library, it checks each piece of information in the attribute value library.

[0113] Software overall architecture as follows Figure 4As shown, the system mainly consists of a rule base construction sub-tool, a semantic model conversion module, a BIM model inspection engine, and a quick modification module. The rule base construction sub-tool converts BIM standards into a domain rule base. For example, the semantic model conversion module converts BIM model files into semantic models. The BIM model inspection engine compares and analyzes the domain rule base and the semantic model. The analysis and review results are exported as a model review report. Finally, the quick modification module provides tools to assist in batch model improvement.

[0114] The main functions of the BIM model automatic review software include:

[0115] (1) Basic settings for model review:

[0116] The model import module is used to import BIM model files into the software. It provides preprocessing functions for model data, such as attribute extraction, to ensure the accuracy and efficiency of subsequent review.

[0117] The basic settings for model inspection can be adjusted according to project requirements. Functions include reviewing model selection, selecting and updating standards and specifications, automatic matching of model files and platform structures, displaying and manually adjusting results, and model filtering.

[0118] The software has a built-in rule base for the project, and the rule base settings include three aspects:

[0119] 1.1 Set the field type:

[0120] Field types include: string, numeric, enumeration, boolean, and time-related. Each type corresponds to certain constraints.

[0121] String: When a string has special formatting specifications, fill in the constraint rule number in 'Formatting Specifications', which corresponds to the specific string inspection requirements.

[0122] Numeric type: Numeric values ​​should meet the constraints of the 'Numerical Range' column.

[0123] Enumeration type: The attribute should be within the range of contents contained in the 'Enumeration value' column.

[0124] Boolean: The value should be either yes or no.

[0125] Date type: The attribute should be in 8-digit number format, in year-month-day order. Hierarchical formats such as "2024-1-1" and month-day-year formats such as "12012023" are incorrect.

[0126] 1.2 Setting up field relationships:

[0127] When data fields have direct relationships, fill in the relevant fields and relationships in "Data Relationships" and "Data Relationship Constraints". For example, the "Top Forward Dimension" field in Table A should be consistent with the 8th field in Table B.

[0128] 1.3 Setting constraints for the same field across different models

[0129] "Data constraints between different models" should be filled in with the same field constraints between different models, such as the location information not being duplicated.

[0130] User-defined rule base: Projects or clients may have different specific requirements. Users can modify the rules on the exported rule base template according to their own needs. The modified template can then be re-imported into the software to perform audits according to the newly defined rules.

[0131] (2) Model review:

[0132] The model review module is one of the core modules for automatic BIM model review. Users can use built-in model checking algorithms to assess the completeness, standardization, logical consistency, and geometric correctness of model information, ensuring that the model meets the requirements of relevant standards and specifications.

[0133] Taking the location information checking algorithm as an example, the steps are as follows:

[0134] 2.1.1 Filter out auxiliary modeling elements in the model file to exclude the influence of irrelevant elements such as two-dimensional lines and text.

[0135] 2.1.2 Extracting 3D model attribute information.

[0136] 2.1.3 According to the grouping specification requirements, filter the construction attributes in the attributes and save the attributes as an external library file for geometric calculation.

[0137] 2.1.4 Extract the positioning information from the construction attributes, and check for standardization, consistency, repeatability, and symbol rules. Save the error results to a report and feed them back to the detailing engineer.

[0138] 2.1.5 For location information that has been verified to be correct, extract the mileage information and calculate the information length of the model using the mileage information.

[0139] 2.1.6 Calculate the actual length of the model using the entity length calculation method.

[0140] 2.1.7 Filter by type and length, summarize and statistically analyze the models that meet the requirements, and provide feedback to the detailing staff.

[0141] 2.1.8 Simultaneously compare the geometric length calculated by the algorithm with the information length calculated by the attribute, and feed back the incorrect model with inconsistent lengths to the depth engineer.

[0142] Taking the calculation of model mileage and offset accuracy as an example, the algorithm is as follows:

[0143] 2.2.1 Create a three-dimensional spatial line for the project center and export the coordinates of the corresponding station points at certain intervals.

[0144] 2.2.2 The program reads the coordinates from the Excel table, generates points, and concatenates them into curve Q1.

[0145] 2.2.3 Project the three-dimensional curve Q1 onto the horizontal plane (Z value is 0) to generate curve Q2.

[0146] 2.2.4 Extract the center point of the calculation model and denote it as P1. Determine the point P2 that P1 is projected onto the horizontal plane (Z value is 0).

[0147] 2.2.5 Determine the point P3 on the Q2 curve that is closest to P2.

[0148] 2.2.6 The mileage marker of the starting point P4 of curve Q2 is ZH1, and the chord length of the distance from P3 to P4 is S. Therefore, the converted mileage marker ZH = ZH1 + S.

[0149] 2.2.7 The distance between P2 and P3 is L, which is the absolute value of the left and right offset values.

[0150] 2.2.8 Determine the tangent vector V1 of curve Q2 at point P3, and the vector V2 pointing from P3 to P2. Determine the angle between V1 and V2 when V1 rotates counterclockwise: when the angle is 90 degrees, the point is to the left with an offset value of L; when the angle is 270 degrees, the point is to the right with an offset value of -L. Review the results based on the final offset values.

[0151] (3) Results presentation:

[0152] Based on the project's built-in BIM standards and specifications, the model undergoes a comprehensive and automated inspection. The results of each review can be categorized and displayed, or saved as an external Excel file, allowing users to trace and review the inspection process, understand the details of each step, and the reasons for the review results.

[0153] (4) Batch modification of models:

[0154] The software currently includes a batch modification module, which provides automatic repair suggestions for detected errors, helping users quickly correct problems in the model. When there are errors in the positioning information, it can automatically calculate and fill in the correct positioning information by comparing it with related attribute information such as IFD code and pier number.

[0155] (5) Review and evaluation:

[0156] Existing technologies lack objective evaluation methods, making it difficult to assess the quality of audits, leading to perfunctory practices and poor audit results.

[0157] To enhance the scientific rigor and effectiveness of the review process, this embodiment innovatively introduces a comprehensive evaluation model for calculating the review accuracy. This method comprehensively considers three key indicators: weight allocation, precision, and recall. Through quantitative analysis, it provides an objective and quantifiable feedback mechanism for review effectiveness.

[0158] This embodiment decomposes the target engineering project into multiple model files, each containing multiple sub-building information models. The overall accuracy rate is calculated as the review result, including:

[0159] Each item in the target audit results is assigned a weight based on the audit difficulty and workload, calculated as follows:

[0160] ;

[0161] in, This represents the weight of the i-th entry. This represents the building difficulty of the i-th entry. This represents the workload of the i-th entry; n represents the number of model files.

[0162] The precision rate for the target project is calculated using the following formula:

[0163] ;

[0164] Where P represents the precision rate of the target project. This represents the precision of the i-th entry in the j-th model file; m represents the total number of reviewed entries.

[0165] The recall rate for the target project is calculated using the following formula:

[0166] ;

[0167] Where R is the recall rate of the target project. This represents the recall rate of the i-th entry in the j-th model file; m represents the total number of reviewed entries.

[0168] Calculate the overall accuracy of the target project The calculation formula is:

[0169] .

[0170] Corresponding to the above method, the present invention also provides an apparatus / system including a computer device, the computer device including a processor and a memory, the memory storing computer instructions, the processor executing the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the apparatus / system performs the steps of the method as described above.

[0171] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned edge computing server deployment method. The computer-readable storage medium can be a tangible storage medium, such as random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, floppy disks, hard disks, removable storage disks, CD-ROMs, or any other form of storage medium known in the art.

[0172] In summary, the BIM model information review method, device, and computer program of this invention, for the cleaned building information model to be reviewed, utilize a pre-trained semantic recognition model to semantically extract the names, functional descriptions, and specification descriptions of each data field in the building information model to be reviewed. The semantics of the data field names are compared with the semantics of the standard field names summarized in the standard attribute field library, and the data fields contained in the standard attribute field library in the building information model to be reviewed are extracted to construct the attribute value library of the model to be reviewed, thus achieving accurate alignment of data fields at the semantic layer. Furthermore, based on the constraints recorded in the attribute value rule library, the numerical values ​​or descriptive text recorded in each data field in the attribute value library of the model to be reviewed are checked, enriching the review items and deepening the constraint methods on numerical values ​​or text in data fields during the review process.

[0173] Furthermore, by configuring corresponding constraint rules for different forms of numerical values ​​within the data fields, we can adapt to the diverse data formats required for BIM model review, thereby improving efficiency and review accuracy.

[0174] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the desired tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave.

[0175] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0176] In this invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.

[0177] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for reviewing BIM model information, characterized in that, The method includes the following steps: Obtain the building information model to be reviewed, which includes geometric and non-geometric information recorded based on data fields; the geometric information includes the location, shape, and size information of the building object; the non-geometric information includes the material, specifications, cost, and construction time information of the building object; The building information model to be reviewed is subjected to data preprocessing, which includes data cleaning and format standardization. Obtain a pre-trained semantic recognition model, and extract the semantics of the data field names in the pre-processed building information model to be reviewed to obtain the corresponding semantics of the field names to be reviewed. Obtain a standard attribute field library for building information model, compare the semantics of the field name to be reviewed with the semantics of the standard field name of each data field in the standard attribute field library, extract the data fields contained in the standard attribute field library of the building information model to be reviewed, and construct an attribute value library of the model to be reviewed. The structure of the attribute value library of the model to be reviewed is the same as that of the standard attribute field library. Obtain the attribute value rule library related to the standard attribute field library, and review the values ​​of each data field in the attribute value library of the model to be reviewed based on the intra-field constraint rules and inter-field relationship constraint rules recorded in the attribute value rule library to obtain the target review result; A summary report on whether the target audit results meet the standards is generated and sent to the target audit port via a preset link; The pre-training steps of the semantic recognition model include: Obtain a training sample set, which includes multiple positive and negative samples. The positive samples contain semantically consistent text pairs within the domain related to the building information model, and the negative samples contain semantically inconsistent text pairs within the domain related to the building information model. Add a label to the positive and negative samples indicating whether they are semantically consistent. An initial neural network model is obtained, which employs a skip-garm model and a similarity calculation module. The skip-garm model performs semantic extraction on text pairs in the positive or negative samples, and the similarity calculation module calculates the similarity of text pairs to obtain the prediction results of consistency comparison. The initial neural network model is trained using the training sample set. The cross-entropy loss function is calculated based on the prediction results and the labels. The parameters of the initial neural network model are updated based on the cross-entropy loss function. The updated skip-garm model is then used as the semantic recognition model. The method is also used to calculate the overall accuracy rate as a review result, including: Each item in the target review results is assigned a weight based on review difficulty and workload, calculated as follows: ； in, This represents the weight of the i-th entry. This indicates the review difficulty of the i-th entry. This represents the workload of the i-th entry; n represents the number of model files; The precision rate for the target project is calculated using the following formula: ； Where P is the precision rate of the target project. This represents the precision of the i-th entry in the j-th model file; m represents the total number of reviewed entries. The recall rate for the target project is calculated using the following formula: Where R is the recall rate of the target project. This represents the recall rate of the i-th entry in the j-th model file; m represents the total number of reviewed entries. Calculate the overall accuracy of the target project. The calculation formula is: 。 2. The BIM model information verification method according to claim 1, characterized in that, The non-geometric information also includes textual description information, which includes functional and specification descriptions of the building. Semantic extraction is performed on the functional description to obtain the semantics of the function to be reviewed, and semantic extraction is performed on the specification description to obtain the semantics of the specification to be reviewed; The consistency of the functional semantics corresponding to the textual description information in the attribute value library of the model to be audited is compared with the standard functional semantics recorded in the standard attribute field library. The consistency of the normative semantics corresponding to the textual description information in the attribute value library of the model to be audited is compared with the standard normative semantics recorded in the standard attribute field library. The target audit result is then expanded based on the comparison results.

3. The BIM model information verification method according to claim 1, characterized in that, The field constraint rules include: Constraint rules within string fields are used to check the string's length, format, value range, uniqueness, whether it is required, and whether there are prohibited characters. Numerical field constraint rules are used to check the range, precision, scale, and unit of the numerical values ​​in the data field. Constraint rules within enumerated fields are used to check whether the enumerated values ​​within the data field are included within a preset range; Boolean field constraint rules are used to check the value form, non-nullability, and dependency on the target field of the Boolean values ​​in the data field. Constraint rules within the date field are used to verify the format and sequence of date data within the data field.

4. The BIM model information verification method according to claim 1, characterized in that, The method further includes: calculating the entity length and entity position of the building information model to be reviewed based on the geometric information, comparing it with the attribute length and attribute position of the standard building information model, generating a location review result, and sending it to the target review port according to the preset link.

5. The BIM model information verification method according to claim 1, characterized in that, The method further includes: when the target audit result contains problematic data fields that do not conform to the field constraint rules and the field relationship constraint rules, searching for other data field contents that generated the problematic data fields, and regenerating and updating the problematic data fields.

6. A BIM model information verification device, comprising a processor, a memory, and a computer program / instructions stored in the memory, characterized in that, The processor is configured to execute the computer program / instructions, and when the computer program / instructions are executed, the device implements the steps of the method as described in any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method as described in any one of claims 1 to 5.

8. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 5.

Images