BIM parameterized design automatic optimization system based on AI

Abstract

The invention relates to the field of BI parameter design, and discloses an AI-based BI parameterization design automatic optimization system, which performs autonomous learning and optimization on BI data through an AI engine fusing reinforcement learning and Bayesian optimization, and automatically generates a design scheme, and the structure, energy consumption, constructability and regulation compliance indexes of the system calculation scheme. According to the method, the problem severity degree is intelligently evaluated, the problem severity degree is divided into a mild problem, a moderate problem and a severe problem, differential processing strategies of parameter fine adjustment, algorithm iteration and scheme regeneration are adopted for different levels, unnecessary global recalculation in a traditional method is avoided, calculation resources are remarkably saved, and the method is suitable for large-scale popularization and application. Meanwhile, the system supports VR immersive man-machine interaction, scheme evaluation and adjustment are facilitated, and the design efficiency, the automation level and the scheme quality are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of BIM parametric design, and more specifically to an AI-based automatic optimization system for BIM parametric design. Background Technology

[0002] Building Information Modeling (BIM) technology, as the core of the digital transformation of the construction industry, has profoundly changed the way projects are designed, constructed, and operated. By creating digital building models that include geometric information and physical and functional characteristics, BIM provides an information-rich collaborative platform for the entire lifecycle management of projects. At the same time, parametric design is increasingly widely used in the field of architectural design. It transforms design intent into adjustable parameters and algorithms by establishing logical connections and constraint rules between design elements. The integration of BIM and parametric design forms parametric BIM, which not only significantly improves the efficiency of creating complex-shaped buildings and the flexibility of design, but also enables rule checks, conflict detection, and performance simulation verification immediately after the scheme is generated through automated verification. When a certain data deviates, the model can be regenerated and optimized according to preset logic. The integration of BIM and parametric design forms parametric BIM, making the creation of complex-shaped buildings, the rapid iteration of design schemes, and the closed-loop feedback of "design-analysis" possible, greatly improving design efficiency and creativity.

[0003] However, the above-mentioned technologies have at least the following technical problems:

[0004] In existing technologies, when deviations in BIM software data parameters are discovered during the verification phase of a design scheme, global re-optimization is often triggered directly. This not only wastes computational resources but also significantly extends the design cycle. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, this invention provides an AI-based automatic optimization system for BIM parametric design to address the problems existing in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] An AI-based automatic optimization system for BIM parametric design includes: a data acquisition module for acquiring BIM software data, which includes geometric and non-geometric data; uniformly encoding the acquired BIM software data to form a multi-source dataset; and outputting the multi-source dataset to an AI optimization engine module; an AI optimization engine module for learning and extending autonomous optimization of the multi-source dataset; employing a hybrid algorithm framework integrating reinforcement learning and Bayesian optimization for hierarchical optimization to generate a BIM software data model that meets the specified conditions; and outputting the BIM software data model to an exponent generation module; and an exponent generation module for receiving the BIM software data model generated by the AI ​​optimization engine module and, through B... The IM software data model calculates structural performance index, energy consumption performance index, constructability index, and regulatory compliance index. The hierarchical feedback module calculates a problem severity index based on these indices, comparing it to a severity threshold to classify problems into three levels: mild, moderate, and severe. Mild problems are corrected through adjustments, moderate problems trigger optimization algorithm iterations, and severe problems trigger a regeneration of the design scheme. The human-computer interaction module receives the adjusted BIM software data parameter model, allowing users to view the BIM model in real-time on a VR terminal. It supports dynamic adjustment of model parameters and scheme evaluation through natural language input, gesture interaction, and immersive operation.

[0008] Preferably, the specific steps for learning, extending, and autonomously optimizing multi-source datasets are as follows: by receiving the processed multi-source datasets from the data acquisition module, learning from the multi-source datasets, reasoning and extending the learned data, and finally autonomously optimizing the design scheme without much human intervention.

[0009] Preferably, the steps for obtaining the structural performance index are as follows: obtaining component identifiers, geometric shapes, spatial locations, and spatial relationships from the geometric data; integrating the component identifiers, geometric shapes, spatial locations, and spatial relationships to obtain component morphology data; obtaining basic attributes, functional and performance data, and management information from the non-geometric data; integrating the basic attributes, functional and performance data, and management information to obtain component attribute data; and calculating the structural performance index using the component morphology data and component attribute data.

[0010] Preferably, the steps for obtaining the energy consumption performance index are as follows: obtaining management and maintenance information from non-geometric data; integrating the management and maintenance information to obtain environmental and operational parameters; obtaining component morphology data, component attribute data, and environmental and operational parameters; and calculating the energy consumption performance index using the component morphology data, component attribute data, and operational parameters.

[0011] Preferably, the steps for obtaining the constructability index are as follows: obtaining geometric shape, spatial location, and spatial relationship information from the geometric data; integrating the geometric shape, spatial location, and spatial relationship information to obtain construction environment data; obtaining management information, basic attributes, and functional and performance data from the non-geometric data; integrating the management information, basic attributes, and functional and performance data to obtain construction element data; obtaining the construction environment data and construction element data; and calculating the constructability index using the construction environment data and construction element data.

[0012] Preferably, the specific steps for obtaining the regulatory compliance index are as follows: acquiring the geometric shape, spatial location, and spatial relationship from the geometric data; integrating the geometric shape, spatial location, and spatial relationship to obtain regulatory-related geometric data; acquiring the basic attributes, type attributes, instance attributes, and functional and performance data from the non-geometric data; integrating the basic attributes, type attributes, instance attributes, and functional and performance data to obtain regulatory-related attribute data; acquiring the regulatory-related geometric data and regulatory-related attribute data; and calculating the regulatory compliance index using the regulatory-related geometric data and regulatory-related attribute data.

[0013] Preferably, the specific steps for the problem severity index are as follows: obtain the component performance index, energy consumption performance index, constructability index, and regulatory compliance index, and calculate the problem severity index by weighted summation of the component performance index, energy consumption performance index, constructability index, and regulatory compliance index.

[0014] Preferably, the specific steps for comparing the problem severity index with the severity threshold and classifying it into three levels—mild, moderate, and severe—are as follows: obtain the problem severity index, compare the problem severity index with the preset severity threshold, when the problem severity index is less than the preset severity threshold, it is a small-scale performance deviation that can be resolved through adjustment, when the problem severity index is equal to the preset severity threshold, it is a large performance deviation with insufficient local correction, requiring the triggering of optimization algorithm iteration, when the problem severity index is greater than the preset severity threshold, it is a systemic deviation or regulatory non-compliance, requiring the regeneration of the design scheme.

[0015] The technical effects and advantages of this invention are as follows:

[0016] This invention significantly improves the automation level and data processing reliability of BIM parametric design by introducing an AI-driven hierarchical optimization and intelligent verification mechanism. The system employs a hybrid algorithm integrating reinforcement learning and Bayesian optimization to achieve autonomous learning and scheme optimization from multi-source data, and conducts comprehensive evaluation based on four major indices: structure, energy consumption, constructability, and regulatory compliance. Through a graded feedback mechanism based on problem severity indices, the system can intelligently distinguish between minor, medium, and severe problems, and adopt differentiated processing strategies such as parameter fine-tuning, algorithm iteration, or scheme reconstruction to avoid resource waste caused by global recalculation. Simultaneously, combined with a VR human-computer interaction module, it supports immersive dynamic adjustment and evaluation, effectively improving design efficiency and scheme quality, reducing manual intervention costs, and ensuring the accuracy of data processing and the flexibility of system response. Attached Figure Description

[0017] Figure 1 This application provides a structural diagram of an AI-based automatic optimization system for BIM parametric design. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The AI-based BIM parametric design automatic optimization system involved in the present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] This invention provides an AI-based automatic optimization system for BIM parametric design, such as... Figure 1 As shown, it includes the following steps:

[0020] The data acquisition module is used to collect BIM software data, which includes geometric and non-geometric data. The collected BIM software data is uniformly encoded to form a multi-source dataset, which is then output to the AI ​​optimization engine module. The AI ​​optimization engine module generates the BIM software data model. It should be noted that the BIM software data is collected through the native application programming interface of the BIM software.

[0021] Geometric data includes the number of geometric shapes, component identification information, geometric and spatial information, and spatial relationship information;

[0022] Component identification information includes a unique ID and a type ID. It should be noted that the unique ID is a globally unique identifier for each component instance in the BIM model, and the type ID identifies the type of the component.

[0023] Geometric and spatial information includes parameter values, three-dimensional geometric shapes, spatial positions and transformations, and parameterized driven dimensions;

[0024] Spatial relationship information includes connectivity, adjacency, and containment relationships;

[0025] Non-geometric data includes basic attributes, functional and performance data, and management information;

[0026] Basic attributes include component name, component category, type attribute, and instance attribute;

[0027] Functional and performance data include performance sensitivity, structural data, thermal data, acoustic data, and illumination data;

[0028] Structural data includes loads, material strength, and reinforcement configuration;

[0029] Thermal data include heat transfer coefficient, solar heat gain coefficient, and specific heat capacity;

[0030] Illumination data includes material reflectivity and transmittance;

[0031] Management information includes stage information, stability coefficient, historical data information, historical maintenance frequency, and cost information.

[0032] The AI ​​optimization engine module is used to learn and extend autonomously optimize multi-source datasets. It uses a hybrid algorithm framework that integrates reinforcement learning and Bayesian optimization to perform hierarchical optimization, and finally generates a BIM software data model that meets the conditions. The generated BIM software data model is then output to the index generation module.

[0033] In this embodiment, it is necessary to specifically explain the steps for learning, extending, and autonomously optimizing the multi-source dataset:

[0034] By receiving the processed multi-source dataset from the data acquisition module, the AI ​​optimization engine module learns from the multi-source dataset, infers and extends the learned data, and finally autonomously optimizes the design scheme without much human intervention. It should be noted that learning from the multi-source dataset refers to the process by which the AI ​​optimization engine module trains and models the geometric and non-geometric data collected from the BIM software using machine learning methods. The autonomous optimization of the multi-source dataset through learning and extension is an existing technology, and this embodiment does not describe its specific steps in detail.

[0035] In this embodiment, it should be noted that the steps for hierarchical optimization using a hybrid algorithm framework that integrates reinforcement learning and Bayesian optimization are as follows:

[0036] Reinforcement learning is a machine learning method that learns the optimal parameter combination through interaction with the environment and based on performance metrics. Bayesian optimization is another method for optimizing black-box functions. It predicts the target function by building a surrogate model and intelligently selects the next most promising parameter point for evaluation to improve search efficiency. It should be noted that the use of a hybrid algorithm framework that integrates reinforcement learning and Bayesian optimization for hierarchical optimization is an existing technology, and this embodiment does not provide a detailed description of its specific steps.

[0037] The index generation module receives the BIM software data model generated by the AI ​​optimization engine module and calculates the structural performance index, energy consumption performance index, constructability index, and regulatory compliance index from the BIM software data model.

[0038] In this embodiment, it should be specifically explained that the steps for obtaining the structural performance index are as follows:

[0039] Obtain component identifiers, geometric shapes, spatial positions, and spatial relationships from the geometric data, and then integrate these data to obtain component morphology data.

[0040] Obtain basic attribute, functional and performance data, and management information from non-geometric data, and integrate the basic attribute, functional and performance data, and management information to obtain component attribute data;

[0041] The structural performance index is calculated using component morphology data and component attribute data. The specific steps are as follows:

[0042]

[0043] In the formula, S is the structural performance index, and G... k D represents the component shape data of the k-th geometric shape. k Let N be the attribute data of the k-th component, and N be the total number of geometric shapes. It should be noted that the structural performance index derivation method based on BIM attributes is used to realize the automated conversion from information integration to data structure performance, which greatly improves efficiency.

[0044] In this embodiment, it should be specifically explained that the steps for obtaining the energy efficiency performance index are as follows:

[0045] Acquire management and maintenance information from non-geometric data, and integrate the management and maintenance information to obtain environmental and operational parameters;

[0046] The energy efficiency index is calculated by acquiring component morphology data, component attribute data, and environmental and operational parameters. The specific steps are as follows:

[0047]

[0048] In the formula, E is the energy performance index, m is the number of energy-related cost parameters, Hc is the number of historical maintenance records, and MD is the energy efficiency index. j For the j-th component's shape data, CD j For the attribute data of the j-th component, P j For the j-th environmental and operational parameter, it should be noted that the number of energy-related cost parameters is obtained from management information in non-geometric data and calculated using the energy performance index calculation algorithm formula. This constructs a computable digital evaluation model of energy performance based on multi-source BIM data, which can intuitively quantify building energy efficiency.

[0049] In this embodiment, it should be specifically explained that the steps for obtaining the constructability index are as follows:

[0050] Obtain information on geometric shapes, spatial locations, and spatial relationships from geometric data, and integrate this information to obtain construction environment data;

[0051] Acquire management information, basic attributes, and functional and performance data from non-geometric data, and integrate the management information, basic attributes, and functional and performance data to obtain construction element data;

[0052] The specific steps for obtaining construction environment data and construction element data are as follows:

[0053]

[0054] In the formula, CL is the constructability index, and C f For the f-th construction environment data, E f Let f be the f-th construction element data, and i be the number of rules for the constructability index. It should be noted that the number of rules for the constructability index is obtained from the construction element data. The advantage of the constructability index calculation formula is that it can perform a comprehensive automated scan and diagnosis of the design scheme by processing construction environment data and construction element data in parallel. It can simultaneously check the compliance of many dimensions such as spatial interference, installation path, construction sequence, process adaptability, and ergonomics, replacing a lot of tedious and error-prone manual review work, improving evaluation efficiency and automation level, and reducing costs and increasing efficiency.

[0055] In this embodiment, it should be specifically explained that the specific steps for obtaining the regulatory compliance index are as follows:

[0056] Obtain geometric shapes, spatial locations, and spatial relationships from geometric data, and integrate these geometric shapes, spatial locations, and spatial relationships to obtain regulatory-related geometric data;

[0057] Obtain basic attributes, type attributes, instance attributes, and functional and performance data from non-geometric data, and integrate these data to obtain regulatory-related attribute data.

[0058] The specific steps for obtaining regulatory-related geometric data and attribute data, and calculating the regulatory compliance index using these data are as follows:

[0059]

[0060] In the formula, RC is the regulatory compliance index, and LD is the regulatory compliance index. a For the a-th regulation-related geometric data, RD a Let 'a' be the regulatory-related attribute data and 't' be the total number of regulations and rules involved in the current inspection. It should be noted that 't' is attribute data in geometric data. The method uses a set theory-based compliance comparison formula, which transforms textual building regulations into automatically executable computer rules and performs batch and rapid compliance comparison and quantitative assessment of BIM design models, achieving comprehensiveness and eliminating omissions, thus reducing risks.

[0061] The graded feedback module calculates the severity index of problems using structural performance index, energy consumption performance index, constructability index, and regulatory compliance index. It compares the severity index with the severity threshold and classifies the problems into three levels: mild, moderate, and severe. Mild problems are corrected through adjustments, moderate problems trigger optimization algorithm iterations, and severe problems trigger the regeneration of design schemes.

[0062] In this embodiment, it should be specifically explained that the specific steps for obtaining the problem severity index are as follows:

[0063] Obtain the component performance index, energy consumption performance index, constructability index, and regulatory compliance index. Calculate the problem severity index by weighted summation of these indices. The specific steps are as follows:

[0064] PT=α×S+β×E+η×CL+χ×RC;

[0065] In the formula, PT is the problem severity index, S is the component performance index, E is the energy consumption performance index, CL is the constructability index, RC is the regulatory compliance index, and α, β, η, and χ are the weighting coefficients of the component performance index, energy consumption performance index, constructability index, and regulatory compliance index. α, β, η, and χ are obtained through an objective weighting method, which determines the weights based on the inherent relationship and degree of variation among the data of each indicator without human intervention. The sum of all weighting coefficients is 1. Using a weighted summation calculation formula, different types of data such as structural performance, energy consumption performance, constructability, and regulatory compliance are uniformly integrated into the problem severity index, avoiding the dominance of a single indicator and ensuring the comprehensiveness of the assessment.

[0066] In this embodiment, it should be specifically explained that the specific steps for comparing the problem severity index with the severity threshold and classifying it into three levels—mild, moderate, and severe—are as follows:

[0067] Obtain the problem severity index and compare it with a preset severity threshold. When the problem severity index is less than the preset severity threshold, it indicates a small-scale performance deviation, which can be resolved through adjustments. When the problem severity index is equal to the preset severity threshold, it indicates a large performance deviation and insufficient local correction, requiring the triggering of optimization algorithm iterations. When the problem severity index is greater than the preset severity threshold, it indicates a systemic deviation or regulatory non-compliance, requiring the regeneration of the design scheme. It should be noted that the preset severity threshold is obtained from historical data information in the management information.

[0068] In this embodiment, it should be noted that the specific steps for obtaining minor issues through adjustment and correction are as follows:

[0069] Obtain parameter values ​​from geometric and spatial information, performance sensitivity from functional and performance data, and stability coefficients from management information. Calculate parameter adjustments and corrections using these parameters, performance sensitivity, and stability coefficients. The specific steps are as follows:

[0070] p n =p c +Δp=p c -η×(M t -M c );

[0071] In the formula, p n Parameter adjustment and correction, p c Here are the parameter values: Δp is the performance sensitivity, η is the stability coefficient, and M... t M is the target value for the performance metric. cThe current indicator value is provided by the national design code. The current indicator value is calculated by the AI ​​optimization engine in the index generation module based on the BIM software data model. The calculation uses a local parameter adjustment correction formula, which couples the parameter adjustment amount with the current performance deviation from the target and its sensitivity to performance, thus achieving precise and adaptive automated fine-tuning.

[0072] In this embodiment, it should be noted that the specific steps for obtaining the optimization algorithm iteration triggered by the moderate problem are as follows:

[0073] When the hierarchical feedback module determines that it is a moderate problem, it indicates that there is a significant defect in the design scheme, triggering the optimization algorithm iteration. Through the AI ​​optimization engine module, a hybrid algorithm framework that integrates reinforcement learning and Bayesian optimization is used to re-optimize the BIM software data parameter model, generating a new scheme. The hierarchical feedback module evaluates the generated new scheme until a BIM software data parameter model that meets the conditions is generated.

[0074] In this embodiment, it should be noted that the specific steps for obtaining the regenerated design scheme triggered by severe problems are as follows:

[0075] When the hierarchical feedback module determines that the problem is a severe problem, it indicates a systemic deviation or regulatory non-compliance. The current design solution has fundamental problems that cannot be solved through optimization algorithm iteration. Therefore, the processing step is to abandon the current solution and regenerate it.

[0076] The human-computer interaction module is used to receive the adjusted BIM software data parameter model, view the BIM model in real time in the VR terminal, and support users to dynamically adjust the model parameters and evaluate the scheme through natural language input, gesture interaction and immersive operation.

[0077] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. 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.

[0078] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An AI-based BIM parametric design automatic optimization system, characterized in that, The system comprises; a data acquisition module for acquiring BIM software data, including geometric data and non-geometric data, uniformly encoding the acquired BIM software data to form a multi-source data set, and outputting the multi-source data set to an AI optimization engine module; the AI optimization engine module is used for learning, extending and autonomously optimizing the multi-source data set, using a hybrid algorithm framework combining reinforcement learning and Bayesian optimization for hierarchical optimization, generating a BIM software data model that meets the conditions, and outputting the BIM software data model to an index generation module; the index generation module is used for receiving the BIM software data model generated by the AI optimization engine module, calculating the structure performance index, energy consumption performance index, constructability index and regulatory compliance index from the BIM software data model; a hierarchical feedback module for calculating a problem severity index from the structure performance index, energy consumption performance index, constructability index and regulatory compliance index, comparing the problem severity index with a severity threshold, and dividing it into three levels: mild, moderate and severe; wherein the mild problem is adjusted and corrected, the moderate problem triggers an optimization algorithm iteration, and the severe problem triggers a new design scheme; a human-computer interaction module for receiving the adjusted BIM software data parameter model, viewing the BIM model in real time in a VR terminal, and supporting users to dynamically adjust and evaluate the model parameters through natural language input, gesture interaction and immersive operation.

2. The AI-based BIM parameterized design automatic optimization system according to claim 1, wherein, The specific steps of learning, extending and autonomously optimizing the multi-source data set are: By receiving the multi-source data set processed in the data acquisition module, learning the multi-source data set, reasoning and extending the learned data, and autonomously optimizing the design scheme without excessive human intervention.

3. The AI-based BIM parameterized design automatic optimization system according to claim 1, wherein, The steps for obtaining the structure performance index are: Obtain the component identification, geometric shape, spatial position and spatial relationship in the geometric data, and obtain the component form data by integrating the component identification, geometric shape, spatial position and spatial relationship; Obtain the basic attributes, function and performance data, and management information in the non-geometric data, and obtain the component attribute data by integrating the basic attributes, function and performance data, and management information; Calculate the structure performance index from the component form data and component attribute data.

4. The AI-based BIM parameterized design automatic optimization system of claim 1, wherein: The steps for obtaining the energy consumption performance index are: Obtain the management information and maintenance information in the non-geometric data, and obtain the environment and operation parameters by integrating the management information and maintenance information; Obtain the component form data, component attribute data and environment and operation parameters, and calculate the energy consumption performance index from the component form data, component attribute data and operation parameters.

5. The AI-based BIM parameterized design automatic optimization system according to claim 1, wherein: The steps for obtaining the constructability index are: Obtain the geometric shape, spatial position and spatial relationship information in the geometric data, and obtain the construction environment data by integrating the geometric shape, spatial position and spatial relationship information; Obtain the management information, basic attributes and function and performance data in the non-geometric data, and obtain the construction element data by integrating the management information, basic attributes and function and performance data; The construction environment data and the construction element data are acquired, and the constructability index is calculated based on the construction environment data and the construction element data.

6. The AI-based BIM parameterized design automatic optimization system according to claim 1, wherein: The specific steps of the regulation compliance index are as follows: The geometric shapes, spatial positions, and spatial relationships in the geometric data are acquired, and the regulation-related geometric data is obtained by integrating the geometric shapes, spatial positions, and spatial relationships; The basic attributes, type attributes, instance attributes, and function and performance data in the non-geometric data are acquired, and the regulation-related attribute data is obtained by integrating the basic attributes, type attributes, instance attributes, and function and performance data; The regulation-related geometric data and the regulation-related attribute data are acquired, and the regulation compliance index is calculated based on the regulation-related geometric data and the regulation-related attribute data.

7. The AI-based BIM parameterized design automatic optimization system of claim 1, wherein: The specific steps of the problem severity index are as follows: The component performance index, the energy consumption performance index, the constructability index, and the regulation compliance index are acquired, and the problem severity index is calculated by weighted summation of the component performance index, the energy consumption performance index, the constructability index, and the regulation compliance index.

8. The AI-based BIM parameterized design automatic optimization system of claim 1, wherein: The specific steps of comparing the problem severity index with the severity threshold to divide the problem severity index into three levels of mild, moderate, and severe are as follows: The problem severity index is acquired, and the problem severity index is compared with the preset severity threshold. When the problem severity index is less than the preset severity threshold, it is a small-range performance deviation, which is solved by adjustment. When the problem severity index is equal to the preset severity threshold, it is a relatively large performance deviation, and local correction is insufficient, so the optimization algorithm iteration needs to be triggered. When the problem severity index is greater than the preset severity threshold, it is a systematic deviation or a regulation non-compliance, and a new design scheme is generated.

Images