Automatic identification and repair method of geometric defects in building BIM data exchange process based on AI technology

Abstract

The present invention discloses a method for automatically identifying and repairing geometric defects in the process of building BIM data exchange based on AI technology, which includes two stages: model training and model use; the logical process of the model training stage includes data collection, data analysis, data conversion, data labeling, data set division, model selection, model architecture design, training model, model evaluation, model deployment and model use; the process of the model use stage includes data input, geometric defect identification, intelligent feedback, geometric repair and data output. The present invention belongs to the field of BIM data conversion technology, and specifically provides a method for automatically identifying and repairing geometric defects in the process of building BIM data exchange based on AI technology. The method can realize intelligent geometric defect detection and identification, and intelligent geometric repair, so as to solve the problems of time-consuming manual repair, low repair efficiency, insufficient precision and low degree of automation existing in the prior art.

Description

Technical Field

[0001] The present invention belongs to the technical field of BIM data conversion, and specifically provides a method for automatically identifying and repairing geometric defects in the construction BIM data exchange process based on AI technology. Background Art

[0002] The data conversion module is an indispensable module in the development process of BIM modeling software. The data conversion module first reads the model file or intermediate format file output by other BIM software, then parses the data in the file, and writes the parsed data into the data structure or format file of its own software.

[0003] During the data exchange process of BIM model files, since the BIM software used by users is different, the internal architecture of the software is different, and the standards for the accuracy requirements are also different, the parsed external data and the data inside the software will be different in terms of geometric topology structure, accuracy, etc. This will inevitably lead to data loss, data parsing errors, data errors and other problems, resulting in the generation of geometric defects during the data exchange process.

[0004] Traditional algorithms for solving geometric defect problems require pre-determining the defect type, traversing the data to identify the defect type, and then repairing it according to a specific geometric model repair algorithm. Defect identification requires human judgment, and the repair process also needs to be performed manually. In actual engineering applications, data exchange between BIM software systems is frequent. Traditional geometric defect repair methods rely on manual operations and limited automated tools. These methods are inefficient, inflexible, and unable to address potential subtle issues in complex models. In addition, as the complexity of BIM models increases, existing repair tools struggle to meet the needs of efficient and accurate repair. Summary of the Invention

[0005] In view of the above situation, in order to overcome the defects of the existing technology, the present invention provides an automatic identification and repair method for geometric defects in the building BIM data exchange process based on AI technology, which realizes intelligent geometric defect detection and identification, and intelligent geometric repair, so as to solve the problems of time-consuming manual repair, low repair efficiency, insufficient precision and low degree of automation in the existing technology.

[0006] The present invention provides the following technical solutions: The present invention proposes an AI-based method for automatically identifying and repairing geometric defects in the building BIM data exchange process, which includes two stages: model training and model use;

[0007] The logical process of the model training phase includes data collection, data analysis, data conversion, data labeling, data set division, model selection, model architecture design, model training, model evaluation, model deployment and model use. The model evaluation step requires model optimization for models that do not meet the evaluation requirements.

[0008] The process of the model usage phase includes data input, geometric defect recognition, intelligent feedback, geometric repair and data output.

[0009] Furthermore, the logical process of the model training phase is as follows:

[0010] (1) Data collection:

[0011] Collect a large amount of BIM model data as initial data and pre-process it to train the model data set;

[0012] (2) Data analysis:

[0013] Receive BIM model data and parse out the topological and geometric information, including fixed points, edges, faces, and surfaces;

[0014] (3) Data conversion:

[0015] The parsed geometric data is converted into a unified internal data format according to the software data conversion module. The internal geometric data format corresponding to this solution is Brep;

[0016] (4) Data annotation:

[0017] The defect types in the converted model are manually identified and annotated in the data. The annotated defect types should be comprehensive.

[0018] (5) Dataset division:

[0019] The labeled dataset is divided into training set, validation set and test set. The training set is used to train the model, the validation set is used to evaluate the model performance and perform hyperparameter tuning, and the test set is used for performance evaluation after the model is finalized.

[0020] (6) Model selection:

[0021] After the data set is prepared, the model training phase begins. Based on the characteristics of Brep's topological structure, the graph neural network (GNN) is selected as the training model.

[0022] Geometry represented by Brep (Boundary Representation) is usually composed of faces, edges, and vertices. The topological relationship of geometric elements can be naturally represented as a graph structure. Vertices and faces can be represented as nodes in the graph, and edges can be represented as edges of the graph. The coordinates, curvature, and number of connected edges of the vertex can be used as feature representations of the vertex, the length, normal vector, and curve type of the edge can be used as feature representations of the edge, and the normal vector, area, surface type, and angle between faces of the face can be used as feature representations of the face. Graph neural networks (GNNs) learn the global representation of graph structures by recursively aggregating feature information of neighboring nodes. GNNs can capture the topological and geometric relationships in Brep geometry, and through deep learning, automatically identify possible geometric defects and repair different geometric defects.

[0023] (7) Model architecture design:

[0024] Design the number of layers of the model and the neuron units in each layer, and choose appropriate activation functions and loss functions;

[0025] (8) Training model:

[0026] Pass the preprocessed data into the model for iterative model training;

[0027] (9) Model evaluation:

[0028] After model training is completed, use the validation set and test set to evaluate the model performance and determine whether the model is overfitting or underfitting;

[0029] If the model's performance meets the preset requirements and the validation error has been minimized and stabilized, the model training is considered complete and the model can be used.

[0030] If the model is judged to be overfitting or underfitting, it is necessary to enter the model optimization process to adjust the model parameters or replace more appropriate activation functions and loss functions. After the adjustment, continue iterative training of the model until a model that meets the requirements is trained;

[0031] (10) Model deployment:

[0032] Integrate the trained model into the BIM data conversion platform as part of geometric defect identification and repair;

[0033] (11) Model usage:

[0034] When using the model, you only need to input the converted data into the trained BIM data conversion platform, and input the converted data from the conventional input into the model. The model will output the detection results of geometric defects and provide repair suggestions. Users can choose the results automatically repaired by the model or perform manual repairs.

[0035] Furthermore, in step (1), data generated from different BIM software are collected as the initial data of the BIM model, which covers various building types and structures. The model should contain a variety of geometric elements, and the data should be stored in various formats including IFC, STEP, and DWG.

[0036] Furthermore, the defect types of the data annotation in step (4) include various geometric defects including topological errors, geometric distortions, and incomplete data.

[0037] Furthermore, the process of the model use phase is as follows:

[0038] The data input process receives and parses the geometric data of the BIM model, including various formats such as IFC, STEP, and DWG. The geometric defect recognition process is based on the trained AI model, automatically identifying defects in the input geometric data, providing intelligent feedback after recognition, and outputting recognition results and repair suggestions. In the geometric repair stage, users can choose to use the repair results output by the model to repair the detected defects or choose to repair them manually. After the repair is complete, the BIM model data is saved again and exported according to the format requirements. This process supports multiple output formats.

[0039] The beneficial effects achieved by the present invention using the above structure are as follows: The method for automatically identifying and repairing geometric defects in the building BIM data exchange process based on AI technology proposed in the present invention has the following advantages:

[0040] (1) The present invention utilizes artificial intelligence technology to automatically identify and repair geometric defects that occur during BIM data exchange, significantly reducing the need for human intervention. Compared with traditional manual repair methods, automated processing significantly increases repair speed, reduces human errors, and reduces labor costs. In particular, the efficiency advantage is more obvious when processing complex and large BIM models.

[0041] (2) The present invention can identify and repair geometric defects with high precision, ensuring the consistency and accuracy of the repaired model with the original design, improving the accuracy and reliability of the repair, and avoiding the precision loss and topological errors that may be caused by traditional methods;

[0042] (3) The AI ​​algorithm of the present invention can adaptively process different types of geometric defects and adapt to a variety of BIM software and data formats. It has a wide range of applications. Users do not need to manually adjust the repair strategy for each project or data format. The system can automatically adapt and identify and repair subtle geometric defects that are usually difficult to detect manually. At the same time, it optimizes the data structure, reduces redundancy, avoids data loss, and enhances the integrity and consistency of the model. It ensures that the geometric information of the BIM model remains intact during the data exchange process, avoiding problems caused by data loss or redundancy.

[0043] (4) The present invention utilizes artificial intelligence technology to identify and repair geometric defects during the BIM software data exchange process, thereby realizing intelligent geometric defect detection and identification, and intelligent geometric repair, so as to solve the problems of time-consuming manual repair, low repair efficiency, insufficient precision, and low degree of automation in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0045] Figure 1 Schematic diagram of the overall structure of an embodiment of the present invention;

[0046] Figure 2 Schematic diagram of the structure of the sealing cover according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0048] See Figure 1-2 The method for automatic identification and repair of geometric defects in the building BIM data exchange process based on AI technology provided in this embodiment includes two stages: model training and model use. Among them, the logical process of the model training stage includes data collection, data analysis, data conversion, data labeling, data set division, model selection, model architecture design, model training, model evaluation, model deployment and model use. The model evaluation step requires model optimization for models that do not meet the evaluation requirements. The process of the model use stage includes data input, geometric defect identification, intelligent feedback, geometric repair and data output.

[0049] Specifically, see Figure 1In this embodiment, the logical process of the model training phase is as follows:

[0050] (1) Data collection:

[0051] Collect data generated by different BIM software as initial BIM model data and pre-process it to form a dataset for training the model. The initial BIM model data type should cover various building types and structures. The model should contain multiple geometric elements, and the data should be stored in various formats (such as IFC, STEP, DWG, etc.);

[0052] (2) Data analysis:

[0053] Receive BIM model data and parse out the topological and geometric information, including fixed points, edges, faces, and surfaces;

[0054] (3) Data conversion:

[0055] The parsed geometric data is converted into a unified internal data format according to the software data conversion module. The internal geometric data format corresponding to this solution is Brep;

[0056] Parsing means reading the information in the model file and identifying its semantics. Conversion means re-storing the read and identified data in an internal data format.

[0057] (4) Data annotation:

[0058] The defect types in the converted model are manually identified and annotated in the data. The annotated defect types should be comprehensive, covering various geometric defects such as topological errors, geometric distortion, and incomplete data.

[0059] (5) Dataset division:

[0060] The labeled dataset is divided into training set, validation set and test set. The training set is used to train the model, the validation set is used to evaluate the model performance and perform hyperparameter tuning, and the test set is used for performance evaluation after the model is finalized.

[0061] (6) Model selection:

[0062] After the data set is prepared, the model training phase begins. Based on the characteristics of Brep's topological structure, the graph neural network (GNN) is selected as the training model.

[0063] Geometry represented by Brep (Boundary Representation) is usually composed of faces, edges, and vertices. The topological relationship of geometric elements can be naturally represented as a graph structure. Vertices and faces can be represented as nodes in the graph, and edges can be represented as edges of the graph. The coordinates, curvature, and number of connected edges of the vertex can be used as feature representations of the vertex, the length, normal vector, and curve type of the edge can be used as feature representations of the edge, and the normal vector, area, surface type, and angle between faces of the face can be used as feature representations of the face. Graph neural networks (GNNs) learn the global representation of graph structures by recursively aggregating feature information of neighboring nodes. GNNs can capture the topological and geometric relationships in Brep geometry, and through deep learning, automatically identify possible geometric defects and repair different geometric defects.

[0064] (7) Model architecture design:

[0065] Design the number of layers of the model and the neuron units in each layer, and choose appropriate activation functions and loss functions;

[0066] (8) Training model:

[0067] Pass the preprocessed data into the model for iterative model training;

[0068] (9) Model evaluation:

[0069] After model training is completed, use the validation set and test set to evaluate the model performance and determine whether the model is overfitting or underfitting;

[0070] If the model's performance meets the preset requirements and the validation error has been minimized and stabilized, the model training is considered complete and the model can be used.

[0071] If the model is judged to be overfitting or underfitting, it is necessary to enter the model optimization process to adjust the model parameters or replace more appropriate activation functions and loss functions. After the adjustment, continue iterative training of the model until a model that meets the requirements is trained;

[0072] (10) Model deployment:

[0073] Integrate the trained model into the BIM data conversion platform as part of geometric defect identification and repair;

[0074] (11) Model usage:

[0075] When using the model, you only need to input the converted data into the trained BIM data conversion platform, and input the converted data from the conventional input into the model. The model will output the detection results of geometric defects and provide repair suggestions. Users can choose the results automatically repaired by the model or perform manual repairs.

[0076] Specifically, see Figure 2 In this embodiment, the process of the model use phase is as follows:

[0077] The data input process receives and parses the geometric data of the BIM model, such as IFC, STEP, DWG, etc. The geometric defect recognition process is based on the trained AI model, automatically identifies defects in the input geometric data, provides intelligent feedback after recognition, and outputs recognition results and repair suggestions. In the geometric repair stage, users can choose to use the repair results output by the model to repair the detected defects or choose to repair them manually. After the repair is completed, the BIM model data is saved again and exported according to the format requirements. This process supports multiple output formats.

[0078] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, material, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, material, or apparatus.

[0079] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An AI-based method for automatically identifying and repairing geometric defects in the building BIM data exchange process, characterized in that: It includes two stages: model training and model use; The logical process of the model training phase includes data collection, data analysis, data conversion, data labeling, data set division, model selection, model architecture design, model training, model evaluation, model deployment and model use. The model evaluation step requires model optimization for models that do not meet the evaluation requirements. The process of the model use phase includes data input, geometric defect recognition, intelligent feedback, geometric repair and data output; The logical process of the model training phase is as follows: (1) Data collection Collect a large amount of BIM model data as initial data and pre-process it to train the model data set; (2) Data analysis Receive BIM model data and parse out the topological and geometric information, including fixed points, edges, faces, and surfaces; (3) Data conversion The parsed geometric data is converted into a unified internal data format according to the software data conversion module. The internal geometric data format corresponding to this solution is Brep; (4) Data annotation The defect types in the converted model are manually identified and annotated in the data. The annotated defect types should be comprehensive. (5) Dataset division The labeled dataset is divided into training set, validation set and test set. The training set is used to train the model, the validation set is used to evaluate the model performance and perform hyperparameter tuning, and the test set is used for performance evaluation after the model is finalized. (6) Model selection After the data set is prepared, the model training phase begins. Based on the characteristics of Brep's topological structure, the graph neural network (GNN) is selected as the training model. (7) Model architecture design Design the number of layers of the model and the neuron units in each layer, and choose appropriate activation functions and loss functions; (8) Training the model Pass the preprocessed data into the model for iterative model training; (9) Model evaluation After model training is completed, use the validation set and test set to evaluate the model performance and determine whether the model is overfitting or underfitting; If the model's performance meets the preset requirements and the validation error has been minimized and stabilized, the model training is considered complete and the model usage phase begins. If the model is judged to be overfitting or underfitting, it is necessary to enter the model optimization process to adjust the model parameters or replace more appropriate activation functions and loss functions. After the adjustment, continue iterative training of the model until a model that meets the requirements is trained; (10) Model deployment Integrate the trained model into the BIM data conversion platform as part of geometric defect identification and repair; (11) Model usage When using the model, you only need to input the converted data into the trained BIM data conversion platform, and input the data converted from conventional input into the model. The model will output the detection results of geometric defects and provide repair suggestions. The user can choose the results automatically repaired by the model or perform manual repairs.

2. The method for automatically identifying and repairing geometric defects in the building BIM data exchange process based on AI technology according to claim 1 is characterized in that: In step (1), data generated from different BIM software are collected as the initial data of the BIM model. The types cover various building types and structures. The model should contain a variety of geometric elements, and the data should be stored in various formats including IFC, STEP, and DWG.

3. The method for automatically identifying and repairing geometric defects in the building BIM data exchange process based on AI technology according to claim 1 is characterized in that: The defect types of data annotation in step (4) include various geometric defects including topological errors, geometric distortion, and incomplete data.

4. The method for automatically identifying and repairing geometric defects in the building BIM data exchange process based on AI technology according to claim 1 is characterized in that: The process of the model usage phase is as follows: The data input process receives and parses the geometric data of the BIM model, including various formats such as IFC, STEP, and DWG. The geometric defect recognition process is based on the trained AI model, automatically identifying defects in the input geometric data, providing intelligent feedback after recognition, and outputting recognition results and repair suggestions. In the geometric repair stage, users can choose to use the repair results output by the model to repair the detected defects or choose to repair them manually. After the repair is completed, the BIM model data is saved again and exported according to the format requirements. This process supports multiple output formats.

Images