Building model family parameter generation method and system based on AI driving

Abstract

The invention discloses a building model family parameter generation method and system based on AI driving. The method comprises the steps that a building design gene pool driving platform is called to extract and screen building model family calibration parameters; inputting the calibration parameters into a building boundary to represent a structured potential geometric tree diffusion model to generate a geometric feature parameter subset; generating a verification hybrid algorithm verification parameter through a building specification embedded parameter, and marking a parameter which does not accord with the rule; the mark parameters are adjusted and corrected in combination with historical data; integrating the parameters, and performing secondary diffusion through a diffusion model to generate an intermediate parameter set; and mapping the intermediate parameter as a building model family target parameter. The system comprises six units which are sequentially connected and cooperatively work. According to the method and the system, historical design data can be fully utilized, deep fusion of specification and parameter generation is realized, the geometric relevance of parameters is enhanced, the problems of low efficiency and lagging of specification verification in the prior art are solved, and the efficiency and the accuracy of parameter generation are improved.

Description

Technical Field

[0001] This invention relates to the field of architectural model family parameter generation technology, and in particular to an AI-driven architectural model family parameter generation method and system. Background Technology

[0002] In the current field of architectural design, with the deep penetration of digital technology, the generation of architectural model family parameters has become a crucial link affecting design efficiency and quality. Traditional architectural model family parameter generation largely relies on manual input and adjustment, requiring designers to repeatedly verify parameters based on building codes, component characteristics, and historical design data. This not only consumes significant time but is also prone to human error leading to parameters that do not meet actual needs. Simultaneously, the architectural industry's requirements for the diversity and standardization of design schemes are constantly increasing. Parameter generation methods relying solely on manual input or simple algorithms are insufficient to quickly respond to the parameter generation needs under different architectural styles and standards, and cannot effectively integrate scattered design data and regulatory systems. This results in insufficient coordination and accuracy in architectural model family parameter generation. Therefore, there is an urgent need to leverage AI technology to build efficient and intelligent parameter generation methods and systems to overcome the limitations of traditional methods.

[0003] Existing technologies for generating architectural model family parameters suffer from two significant drawbacks. Firstly, current parameter generation methods lack in-depth utilization of architectural design gene data and fail to establish a unified gene bank-driven mechanism. This makes it difficult to effectively draw upon historical design experience during parameter generation, with parameter adjustments relying heavily on subjective judgment. This lack of a standardized and reusable parameter generation logic leads to significant differences in generation efficiency across different projects and low matching rates between parameters and historically successful design schemes. Secondly, the integration of building codes and parameter generation in existing technologies is insufficient. Code verification is often performed independently after parameter generation, failing to embed code requirements into the entire process. This results in generated parameters frequently failing to meet code standards, requiring repeated adjustments and corrections. This not only increases the complexity of the parameter generation process but may also disrupt the geometric relationships between parameters due to later adjustments, affecting the overall coordination and accuracy of the architectural model family. Summary of the Invention

[0004] To overcome the shortcomings and deficiencies of existing technologies, this invention provides an AI-driven method and system for generating building model family parameters.

[0005] The technical solution adopted in this invention is an AI-driven method for generating architectural model family parameters, comprising: Step S1, calling the Hektar architectural design gene library driving platform to extract a set of basic parameters for the architectural model family, which includes geometric feature parameters of building components, material property parameters, and spatial layout related parameters; using the platform's built-in data filtering module to perform feature filtering on the extracted basic parameter set, retaining calibration parameters directly related to the generation of the architectural model family; Step S2, inputting the calibration parameters filtered in Step S1 into a structured potential geometric tree diffusion model representing the building boundary, constructing a multi-dimensional geometric feature diffusion network; performing hierarchical diffusion processing on the geometric features corresponding to the calibration parameters in this network, generating subsets of geometric feature parameters at different diffusion levels; Step S3, inputting the subsets of geometric feature parameters generated in Step S2 into a hybrid algorithm for parameter generation and verification embedded in building codes, which first constructs a parameter verification rule base based on a preset building code system, and then uses a rule matching module... Step S4: Match the subset of geometric feature parameters one by one with the rules in the verification rule base, and mark the parameters that do not conform to the rules. Step S5: Adjust the parameters that do not conform to the rules marked in Step S3. Based on the parameter optimization module in the hybrid algorithm for parameter generation and verification embedded in the building code, and combined with the historical optimization data stored in the Hektar building design gene library driving platform, generate a parameter adjustment scheme and correct the parameters that do not conform to the rules according to the scheme. Step S6: Integrate the parameters corrected in Step S4 with the parameters that conform to the rules to form the initial parameter set of the building model family. Input the initial parameter set into the building boundary representation structured potential geometric tree diffusion model for secondary diffusion processing to strengthen the geometric correlation between parameters and generate the intermediate parameter set of the building model family. Step S7: Perform parameter mapping processing on the intermediate parameter set generated in Step S5. Through the building model family parameter generation module, map the intermediate parameter set into target parameters that the building model family can recognize, and complete the generation of building model family parameters.

[0006] Furthermore, the expression for the structured potential geometric tree diffusion model representing the building boundary is: ,in, This represents the post-diffusion geometric feature parameters output by the model. These are the calibration parameters selected in step S1. The number of diffusion levels. The diffusion coefficient of the model is . The number of layers for geometric features, For the first Weighting coefficients of layer geometric features For the first Layer geometric feature diffusion function, For activation function, For the first The weight matrix of the layer; Including building component length parameters Width parameters and height parameters The value range is 1-10. The value range is 0.1-0.8. satisfy for Dimensional matrix, To calibrate the number of parameters, For the first Number of layer output parameters.

[0007] Furthermore, the expression for the hybrid algorithm for generating and validating embedded parameters in the building code is as follows: ,in, This represents the parameter verification result output by the algorithm. The subset of geometric feature parameters generated in step S2, For the set of rules in the parameter validation rule base, This is a historical data set within the Hektar architectural design gene bank-driven platform. To verify the number of rules, For the first The validation weight of each rule For the first Validation rule function, For data fusion operators, In order to be with the first The historical data corresponding to the rule; Including component position parameters and angle parameters , Used for judgment Does it comply with building spacing specifications and component size specifications? The range of values ​​is This includes the pass rate and adjustment range data for similar parameters in the past.

[0008] Furthermore, the data interaction expression between the Hektar architectural design gene bank driving platform and the architectural model family parameter generation module is as follows: ,in, This represents the data after the interaction. For gene bank data in the platform, The intermediate parameter set generated in step S5 For data interaction compensation coefficient, For data fusion weights, For compensation coefficient, For data convolution operators; the Including architectural style parameters and component combination parameters, This includes parameter correlation and parameter accuracy data. The range of values ​​is The range of values ​​is Adjustments are made dynamically based on data interaction latency and data volume.

[0009] Furthermore, the parameter mapping expression of the building model family parameter generation module is as follows: ,in, Indicates the target parameters after mapping. For the intermediate parameter set, For a set of parameter mapping rules, For a set of mapping vectors, For the number of mapping rules, For the first The weight of each mapping rule, For the first Mapping rule function, For the first The mapping vector corresponding to each rule; This includes parameter priority and parameter compatibility data. Used to convert intermediate parameters into component parameters and attribute parameters for building model families. The value range is 0.2-0.8. for Dimensional vector, For the target parameter dimension.

[0010] Furthermore, the collaborative expression of the building boundary representation structured potential geometric tree diffusion model and the building code embedded parameter generation and verification hybrid algorithm is as follows: ,in, This indicates the output of collaborative work. For the output of the diffusion model, For the output of the verification algorithm, This is the correction value for the cooperative error. For collaborative weights; the Including diffusion parameter deviations and diffusion efficiency data, This includes data on verification accuracy and parameter adjustment suggestions. The range of values ​​is Dynamic calculation based on data deviations during the collaboration process.

[0011] Further, step S3 specifically includes the following sub-steps: Step S31: Retrieve a preset building code system from the storage module of the hybrid algorithm for generating and verifying embedded parameters in building codes. This system includes building structural safety codes, building fire protection codes, and building space use codes. Decompose each code into specific parameter constraints to form a list of parameter constraint entries; Step S32: Classify the geometric feature parameter subsets generated in step S2. Divide the parameter subsets into wall parameter subsets, floor slab parameter subsets, and beam and column parameter subsets according to the building component type. Label each subset with the corresponding component identification information; Step S33: Associate the classified parameter subsets with the list of parameter constraint entries formed in step S31. Compare each parameter in the parameter subset with the requirements of the corresponding constraint entry through the rule matching module, and record the matching results of the parameters and constraint entries; Step S34: Mark the parameters that do not conform to the constraint entries according to the matching results. Add a non-conformity reason label to the marked parameters. The label includes the corresponding constraint entry number and the specific numerical information of the parameter exceeding the range.

[0012] Further, step S4 specifically includes the following sub-steps: Step S41: Extract historical parameter adjustment data of the same type as the marked parameters in step S3 from the historical database of the Hektar architectural design gene bank driven platform. This data includes the initial value of the historical parameters, the adjusted values, and the standard clauses on which the adjustment is based. Step S42: Input the extracted historical adjustment data into the parameter optimization module of the building code embedded parameter generation and verification hybrid algorithm. Analyze the patterns of the historical adjustment data through the data mining unit in the module to generate a parameter adjustment trend curve. Step S43: Based on the adjustment trend curve and the non-compliance reason labels of the marked parameters in step S3, construct a parameter adjustment objective function. This function aims to ensure that the parameters meet the standard requirements and are close to the historical optimal adjustment value. Step S44: Calculate the parameter adjustment range and direction based on the parameter adjustment objective function to form a specific parameter adjustment plan. The plan clearly defines the values ​​of each marked parameter before and after adjustment.

[0013] Further, step S5 specifically includes the following sub-steps: Step S51: Collect the parameters corrected in step S4 and the parameters that conform to the rules in step S3, and perform format unification processing on the two types of parameters, adjusting the numerical format and unit expression (if any) of the parameters to a consistent standard to form a preliminary parameter set; Step S52: Perform parameter correlation analysis on the preliminary parameter set, calculate the correlation coefficient between parameters through the correlation analysis unit in the building boundary representation structured potential geometric tree diffusion model, and identify parameter groups with strong correlation; Step S53: Input the preliminary parameter set into the building boundary representation structured potential geometric tree diffusion model, set the number of levels and diffusion coefficients for secondary diffusion, start the model to perform secondary diffusion processing, and strengthen the collaborative diffusion of parameter groups with strong correlation during the diffusion process; Step S54: Integrate the parameters after secondary diffusion processing, remove redundant parameters generated during the diffusion process, retain the calibrated valid parameters, and form an intermediate parameter set for the building model family.

[0014] The AI-driven architectural model family parameter generation system includes: a Hektar architectural design gene bank parameter extraction unit, connected to the architectural boundary representation structured latent geometric tree diffusion model unit, used to extract the basic parameter set of the architectural model family from the Hektar architectural design gene bank driving platform and perform feature filtering, transmitting the filtered calibration parameters to the architectural boundary representation structured latent geometric tree diffusion model unit; an architectural boundary representation structured latent geometric tree diffusion model unit, connected to both the Hektar architectural design gene bank parameter extraction unit and the building code embedded parameter verification unit, used to perform hierarchical diffusion processing on the calibration parameters to generate a subset of geometric feature parameters, and transmit this subset to the building code embedded parameter verification unit; and a building code embedded parameter verification unit, connected to both the architectural boundary representation structured latent geometric tree diffusion model unit and the parameter optimization and adjustment unit, used for... According to building codes, a subset of geometric feature parameters is validated and parameters that do not conform to the rules are marked. The marked parameters are then transmitted to the parameter optimization and adjustment unit. The parameter optimization and adjustment unit is connected to the building code-embedded parameter validation unit and the parameter integration secondary diffusion unit. It is used to generate parameter adjustment schemes based on historical data and correct the marked parameters. The corrected parameters are then transmitted to the parameter integration secondary diffusion unit. The parameter integration secondary diffusion unit is connected to the parameter optimization and adjustment unit and the building model family target parameter generation unit. It is used to integrate the parameters that conform to the rules and the corrected parameters, input the building boundary representation structured potential geometric tree diffusion model for secondary diffusion, generate an intermediate parameter set, and transmit it to the building model family target parameter generation unit. The building model family target parameter generation unit is connected to the parameter integration secondary diffusion unit. It is used to perform parameter mapping on the intermediate parameter set and generate target parameters that can be recognized by the building model family.

[0015] Beneficial Effects: This invention proposes an AI-driven method and system for generating architectural model family parameters. By calling a dedicated architectural design gene bank-driven platform to extract and filter basic parameters, a unified gene bank-driven mechanism is constructed. This changes the current situation where existing technologies lack in-depth utilization of design gene data, allowing parameter generation to fully draw on historical design experience, forming a standardized and reusable parameter generation logic, reducing the efficiency differences in parameter generation between different projects, and improving the matching degree between parameters and historical high-quality design schemes. At the same time, a hybrid algorithm for parameter generation and verification embedded in building codes is integrated into the entire parameter generation process. Verification is performed simultaneously according to building codes during parameter generation, rather than independently after generation, avoiding the generation of parameters that do not conform to the codes, reducing repeated adjustment and correction steps, and lowering the complexity of the parameter generation process. Furthermore, by using a structured potential geometric tree diffusion model representing building boundaries to perform hierarchical diffusion and secondary diffusion processing on parameters, the geometric correlation between parameters is strengthened, preventing the destruction of correlation in later adjustments, ensuring the overall coordination and accuracy of the architectural model family, significantly improving the efficiency and accuracy of parameter generation, and meeting the architectural industry's needs for diverse and standardized design schemes. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method steps of the present invention;

[0017] Figure 2 This is a diagram showing the system unit composition of the present invention. Detailed Implementation

[0018] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] like Figure 1 As shown, the AI-driven method for generating building model family parameters includes:

[0020] Step S1: Call the Hektar architectural design gene library driving platform to extract the basic parameter set of the architectural model family. This set includes geometric feature parameters of architectural components, material property parameters and spatial layout related parameters. The extracted basic parameter set is filtered by the data filtering module built into the platform to retain the calibration parameters that are directly related to the generation of the architectural model family.

[0021] Specifically, step S1 utilizes a professional architectural design gene bank-driven platform to filter calibration parameters directly related to the generation of the architectural model family from massive amounts of data. Redundant information is eliminated, providing high-quality data support for subsequent model calculations and algorithm verification. This avoids invalid data consuming computational resources and reducing process efficiency, while ensuring the parameters are specific and accurately reflect the key attributes of building components. This lays a reliable foundation for subsequent geometric feature diffusion and specification verification, guaranteeing that the parameters of the final generated architectural model family meet actual design requirements. This step achieves precise parameter selection through the platform's built-in filtering module, solving the problems of low efficiency and strong subjectivity associated with traditional manual filtering, and improving the standardization and accuracy of parameter acquisition. In the specific implementation process, the data extraction function of the Hektar architectural design gene bank driving platform is first activated. The extraction scope is set as the basic parameters of the high-rise residential building model family. The extracted basic parameter set includes three categories: geometric feature parameters of building components (including wall length, floor slab width, beam and column height, etc., where the wall length parameter ranges from 2400 to 7200, the floor slab width parameter ranges from 3000 to 6000, and the beam and column height parameter ranges from 2800 to 3300), material property parameters (including concrete density, steel reinforcement strength, etc., where the concrete density parameter is fixed at 2400, and the steel reinforcement strength parameter ranges from 400 to 500), and spatial layout related parameters (including component spacing, relative position coordinates, etc., where the component spacing parameter ranges from 300 to 800, and the relative position coordinate parameter takes the building reference point as the origin, with the X-axis value ranging from 0 to 10000 and the Y-axis value ranging from 0 to 8000). After extraction, the filtering module is activated, and the filtering condition is set to "directly related to component forming, assembly and structural stability". The module calculates the matching degree between parameters and conditions, and retains parameters with a matching degree of more than 90% as calibration parameters. In the final set of calibration parameters, geometric feature parameters account for 60%, material property parameters account for 25%, and spatial layout related parameters account for 15%, and the missing rate of each type of parameter is less than 2%, ensuring that the calibration parameters can fully support the subsequent steps.

[0022] Step S2: Input the calibration parameters filtered in step S1 into the building boundary representation structured potential geometric tree diffusion model to construct a multi-dimensional geometric feature diffusion network. In this network, perform hierarchical diffusion processing on the geometric features corresponding to the calibration parameters to generate subsets of geometric feature parameters under different diffusion levels.

[0023] Specifically, step S2 uses the building boundary to represent a structured potential geometric tree diffusion model, constructing a hierarchical diffusion network. This expands the planar calibration parameters into a subset of geometric feature parameters with hierarchical relationships, enabling the parameters to reflect the geometric form of building components from different precision dimensions. This provides more detailed parameter basis for subsequent standard verification and offers more geometric feature selections for diverse designs of building model families. Simultaneously, hierarchical diffusion strengthens the geometric logic relationship between parameters, avoiding the problems of single geometric feature expression and weak correlation in traditional parameter processing, thus improving the completeness and accuracy of the parameters' description of the component's geometric form. In practice, the calibration parameters selected in S1 are input into the model in the form of a data stream. A multi-dimensional geometric feature diffusion network with three diffusion levels is first constructed within the model: the first layer corresponds to the overall outline accuracy of the component (error allowable range ±2%), the second layer corresponds to the local detail accuracy of the component (error allowable range ±1%), and the third layer corresponds to the interface connection accuracy of the component (error allowable range ±0.5%). After initiating layered diffusion, based on geometric feature parameters, the first layer generates a subset of overall component contour parameters. The wall length parameter, after diffusion, ranges from 2400 to 7200 (deviation controlled within ±2%, i.e., maximum deviation 48-144), and the floor slab width parameter, after diffusion, ranges from 3000 to 6000 (deviation ±60-120). The second layer generates a subset of local detail parameters based on the parameters from the first layer, such as the depth of grooves on the wall surface (range 50-150, accuracy retained to two decimal places) and the chamfer radius of the floor slab edge (range 20-50, accuracy retained to two decimal places). The third layer generates a subset of interface connection parameters based on the parameters from the first two layers, such as the dimensions of beam-column connection nodes (length range 300-500, width range 200-350) and the connection angle between the wall and the floor slab (range 89.5°-90.5°). The final result is three parameter subsets. Each subset contains between 50 and 80 parameters, depending on the component type, and the logical correlation between the parameters within each subset is higher than 85%.

[0024] Step S3: Input the subset of geometric feature parameters at each level generated in step S2 into the hybrid algorithm for parameter generation and verification embedded in the building code. This algorithm first constructs a parameter verification rule base based on the preset building code system, and then uses the rule matching module to match the subset of geometric feature parameters with the rules in the verification rule base one by one, marking the parameters that do not conform to the rules.

[0025] Specifically, step S3 uses a hybrid algorithm for parameter generation and verification embedded in building codes to transform various code requirements, such as those related to building structural safety, fire protection, and space utilization, into executable verification rules. Verification is performed simultaneously during parameter generation, identifying parameters that do not conform to the codes in advance. This avoids large-scale adjustments after parameter generation, reduces rework costs, and ensures that parameters comply with industry standards, meeting the safety and functional requirements of building design. This provides code assurance for the practical application of building model families, solves the problems of fragmented code application and incomplete verification in traditional verification, and improves the systematicness and efficiency of verification. During implementation, the three hierarchical parameter subsets generated by S2 are first imported into the algorithm. The algorithm automatically retrieves a preset parameter verification rule library, which includes 280 specific rules: Building structural safety codes (e.g., parameter limits corresponding to component load-bearing strength, concrete component compressive strength parameter limit not less than 20, steel component tensile strength parameter limit not less than 360), building fire protection codes (e.g., component spacing parameter standards, evacuation passage wall spacing parameter not less than 1200, fire hydrant and wall spacing parameter not less than 500), and building space usage codes (e.g., component space occupancy ratio parameter range, bedroom area floor slab space occupancy ratio not exceeding 70%, living room area beam and column space occupancy ratio not exceeding 15%). Then, the rule matching module is activated, matching each parameter with its corresponding rule one by one in the order of "overall outline first, then local details, and finally interface connection," recording the difference between the actual parameter value and the rule limit, and marking parameters exceeding the limit. The marking information includes the parameter name, its level, rule number, and deviation value (e.g., "wall length - first level - rule 58 - deviation +300"). Ultimately, the marking rate of each parameter subset at each level was controlled between 8% and 12%, with the overall contour level marking rate at approximately 8%, the local detail level marking rate at approximately 10%, and the interface connection level marking rate at approximately 12%, ensuring accurate identification of non-compliant parameters without excessively increasing the burden of subsequent adjustments.

[0026] Step S4: Adjust the parameters that do not conform to the rules marked in step S3. Based on the parameter optimization module in the hybrid algorithm for parameter generation and verification embedded in the building code, and combined with the historical optimization data stored in the Hektar building design gene bank driving platform, generate a parameter adjustment scheme and correct the parameters that do not conform to the rules according to the scheme.

[0027] Specifically, step S4 combines building code requirements with historical optimization data to generate a scientific adjustment plan through the algorithm's parameter optimization module. This ensures that the corrected parameters not only comply with the code but also maintain geometric and logical consistency with other compliant parameters, avoiding imbalances in the overall parameter system caused by adjusting a single parameter. Simultaneously, historical data is used to improve the rationality and efficiency of the adjustment, reduce trial-and-error costs, and ensure the smooth implementation of subsequent parameter integration and secondary diffusion. This addresses the problems of traditional parameter adjustments relying on experience and lacking data support, improving the accuracy and reliability of the adjustment and laying the foundation for generating high-quality parameters. In practice, 150 records of similar non-compliant parameter adjustments from the past 5 years of high-rise residential building model families are retrieved from the historical database of the Hektar architectural design gene bank-driven platform. These records include the initial historical parameter value (e.g., the initial length of a wall was 6800), the adjusted value (7000), the code clauses on which the adjustment was based (e.g., Article 3.2.1 of the "Code for Design of Building Structures" GB50009-2012), and the application effect after adjustment (e.g., a 12% improvement in structural stability). These data are input into the parameter optimization module, which uses statistical analysis to identify patterns: Wall length parameters typically see adjustments of 3%-8% of their original value, usually increasing to meet spacing specifications; beam and column dimensions typically see adjustments of 2%-5% of their original value, with adjustments varying depending on load-bearing requirements. Combining the reasons for non-compliance with S3-marked parameters (e.g., "wall length 6800 < specification limit 7000"), an adjustment objective function is constructed to calculate the adjustment range and direction for each marked parameter. For example, the wall parameter with a length of 6800 is adjusted to 7000 (range +200, increase 2.94%), and the beam and column parameter with a width of 280 is adjusted to 300 (range +20, increase 7.14%). After adjustment, the corrected parameters are validated a second time to ensure a 100% pass rate. Simultaneously, the adjusted data is entered into the historical database to update the adjustment patterns for similar parameters.

[0028] Step S5: Integrate the parameters corrected in step S4 with the parameters that conform to the rules to form an initial parameter set for the building model family. Input the initial parameter set into the building boundary representation structured potential geometric tree diffusion model for secondary diffusion processing to strengthen the geometric correlation between parameters and generate an intermediate parameter set for the building model family.

[0029] Specifically, step S5 integrates compliant parameters and corrected parameters to form a complete initial parameter set. This initial set is then subjected to secondary model diffusion processing to strengthen the geometric relationships between parameters. This ensures that the parameters not only conform to specifications but also maintain geometric consistency, avoiding conflicts between parameters. Simultaneously, missing detailed parameters are supplemented, enhancing parameter richness and addressing the issues of insufficient correlation and missing details after traditional parameter integration. This provides a complete and coordinated intermediate parameter foundation for the subsequent generation of target parameters, ensuring that the final parameters accurately support the construction of the building model family. During implementation, the corrected parameters from S4 (approximately 35, accounting for 15%) and the compliant parameters from S3 (approximately 198, accounting for 85%) are collected. The formats of both types of parameters are standardized: numerical formats are uniformly retained to two decimal places (e.g., 2400 is adjusted to 2400.00, and 350 is adjusted to 350.00), and parameter names are named according to "component type - parameter category - parameter dimension" (e.g., "wall - geometric feature - length" "beam and column - material property - strength"), forming an initial parameter set of 233 parameters with a 100% parameter format uniformity rate. The initial parameter set is then input into the structured potential geometric tree diffusion model representing the building boundary. The secondary diffusion is set to two layers: the first layer is parameter association enhancement diffusion, where the model calculates the geometric association coefficients between parameters. Parameter groups with association coefficients higher than 0.7 (such as wall length and floor slab width, beam / column height and floor slab thickness) undergo collaborative diffusion to ensure synchronized parameter adjustments within the group (e.g., when the wall length increases by 200, the floor slab width increases by 150). The second layer is detailed parameter supplement diffusion, supplementing 28 detailed parameters based on existing parameters, such as component surface flatness (range 0.5-2.0, precision 0.01) and interface adaptation precision (range 0.1-0.8, precision 0.01), accounting for 12% of the total parameters. After the secondary diffusion, 8 duplicate parameters (such as "wall thickness -1" and "wall thickness -2") and 5 parameters with no substantial impact (such as "component surface color number") are removed, ultimately generating an intermediate parameter set containing 248 parameters. The redundancy removal rate is 5.5%, and the parameter association degree is improved to 90%.

[0030] Step S6: Perform parameter mapping processing on the intermediate parameter set generated in step S5. The intermediate parameter set is mapped into target parameters that can be recognized by the building model family through the building model family parameter generation module, thus completing the building model family parameter generation.

[0031] Specifically, step S6 uses the building model family parameter generation module to map the intermediate parameters that have undergone multiple rounds of optimization into target parameters that the model can directly recognize. This transforms the parameters from data form to application form, ensuring that the parameter format and numerical range match the target model and avoiding incompatibility issues. At the same time, final verification ensures the integrity and logic of the parameters, providing directly usable parameters for the design, rendering, and assembly of the building model family. This solves the problems of poor format compatibility and missing verification in traditional parameter conversion, improves the efficiency and reliability of parameter application, and completes the entire parameter generation process. In practice, the 248 intermediate parameters generated by S5 are first imported into the parameter generation module. The module reads the parameter format requirements of the high-rise residential building model family: in terms of data type, geometric dimension parameters are integers (such as length and width), and material property parameters are decimals (such as density and strength, retained to two decimal places); in terms of numerical range, wall length is 3000-8000, floor slab width is 2500-6500, concrete density is 2400.00-2500.00, and steel reinforcement strength is 400.00-500.00; the naming rule must include the component number (such as "wall-W1-length"). Then, the mapping function is started, and the parameters are converted according to the correspondence between "intermediate parameter category" and "target parameter category". For example, "wall-geometric feature-length" is mapped to "wall-W1-length". When the value exceeds the range, it is fine-tuned (such as the intermediate parameter "wall length 8200" is fine-tuned to 8000, with an amplitude of -200 and a fine-tuning rate of 0.8%) to ensure that all parameters meet the format requirements. After mapping, an integrity check is performed to verify whether key parameters (such as wall length, beam and column strength, and component spacing) are missing (56 key parameters were checked, none were missing), and whether there are any conflicts in the logical relationships between parameters (e.g., the wall spacing of 1200mm complies with fire protection specifications, so there are no conflicts). The verification pass rate is 100%. Finally, a set of 248 target parameters is generated, including 85 basic component parameters, 92 geometric feature parameters, and 71 assembly-related parameters. This set can be directly imported into architectural model design software for model building, rendering, and assembly.

[0032] Preferably, the expression representing the structured potential geometric tree diffusion model of the building boundary is: ,in, This represents the post-diffusion geometric feature parameters output by the model. These are the calibration parameters selected in step S1. The number of diffusion levels. The diffusion coefficient of the model is . The number of layers for geometric features, For the first Weighting coefficients of layer geometric features For the first Layer geometric feature diffusion function, For activation function, For the first The weight matrix of the layer; Including building component length parameters Width parameters and height parameters The value range is 1-10. The value range is 0.1-0.8. satisfy for Dimensional matrix, To calibrate the number of parameters, For the first Number of layer output parameters.

[0033] Specifically, the building boundary represents the computational logic and parameter range of the structured potential geometric tree diffusion model. By setting reasonable weight coefficients, diffusion levels, and activation function application rules, the model ensures accurate processing of the calibration parameters after S1 filtering, generating a parameter subset that meets the geometric feature requirements of the building model family. This avoids excessive parameter deviation or incomplete geometric feature representation during diffusion, while providing clear parameter configuration guidelines for standardized model applications, improving the stability and reliability of model computation results, and supporting the efficient implementation of subsequent standard verification steps. During implementation, the specific categories and numerical ranges of calibration parameters must be determined before model computation. For example, the length parameter of building components is between 2400-7200, the width parameter is between 3000-6000, and the height parameter is between 2800-3300. The diffusion level is set to 1-10 layers, specifically selected according to the complexity of the building components. Simple components such as standard walls are selected with 3-5 layers, while complex components such as irregular beams and columns are selected with 8-10 layers. The model diffusion coefficient is controlled between 0.1 and 0.8, gradually decreasing with increasing layer level. The diffusion coefficient for the first layer is set to 0.7-0.8, and for the top layer to 0.1-0.2. The number of geometric feature layers is determined based on the parameter category, with 3-5 layers for geometric feature parameters. The total weight coefficient of each layer is 1. The weight coefficient of the first layer (overall contour features) accounts for 40%-50%, the middle layer (local detail features) accounts for 30%-40%, and the bottom layer (interface connection features) accounts for 10%-20%. The activation function is used to enhance the non-linear expressive power of the parameters. The dimension of the weight matrix is ​​determined based on the number of calibrated parameters and the number of output parameters. If the number of calibrated parameters is 50-80 and the number of output parameters per layer is 30-50, then the weight matrix is ​​set to a dimension of 50-80 rows and 30-50 columns to ensure that the model can generate an accurate subset of geometric feature parameters through multi-dimensional operations.

[0034] Preferably, the expression for the hybrid algorithm for generating and validating embedded parameters in the building code is: ,in, This represents the parameter verification result output by the algorithm. The subset of geometric feature parameters generated in step S2, For the set of rules in the parameter validation rule base, This is a historical data set within the Hektar architectural design gene bank-driven platform. To verify the number of rules, For the first The validation weight of each rule For the first Validation rule function, For data fusion operators, In order to be with the first The historical data corresponding to the rule; Including component position parameters and angle parameters , Used for judgment Does it comply with building spacing specifications and component size specifications? The range of values ​​is This includes the pass rate and adjustment range data for similar parameters in the past.

[0035] Specifically, the hybrid algorithm for parameter generation and verification embedded in building codes, by clearly defining the algorithm's verification logic, parameter values, and data fusion rules, enables the algorithm to comprehensively and accurately verify the subset of geometric feature parameters generated by S2 according to the preset code system. It accurately marks parameters that do not conform to the code, avoiding missed or false detections caused by ambiguous verification rules or unreasonable weight allocation. Simultaneously, it leverages historical data to improve the rationality of the verification results, providing a reliable basis for subsequent parameter adjustments and ensuring the compliance of building model family parameters. During implementation, the algorithm first retrieves the parameter verification rule library, which includes 280 specific rules corresponding to various codes related to building structural safety, fire protection, and space usage. Examples include limits on the compressive strength of concrete components not less than 20, the distance between walls on both sides of evacuation routes not less than 1200, and the floor slab occupying no more than 70% of the space in bedroom areas. During the verification process, the subset of geometric feature parameters includes component position parameters (X-axis 0-10000, Y-axis 0-8000) and angle parameters (89.5°-90.5°). The verification weights are set according to the importance of the rules: 0.7-0.9 for structural safety rules, 0.5-0.7 for fire protection rules, and 0.2-0.5 for space utilization rules. The historical data set includes parameter verification records for similar building model families over the past 5 years. Each record includes the historical parameter value, verification result, adjustment plan, and application effect. For example, a wall length parameter with an initial value of 6800 was marked because it was less than the code limit of 7000; after adjustment to 7000, the structural stability improved by 12%. The data fusion operator is used to integrate the correlation information between parameters and historical data, ensuring that the verification results not only meet current code requirements but also reference historical optimization experience. When marking parameters that do not conform to the rules, the deviation value between the parameter and the code limit must be clearly recorded, calculated to the nearest whole number, providing a clear basis for subsequent parameter adjustments.

[0036] Preferably, the data interaction expression between the Hektar architectural design gene bank driving platform and the architectural model family parameter generation module is as follows: ,in, This represents the data after the interaction. For gene bank data in the platform, The intermediate parameter set generated in step S5 For data interaction compensation coefficient, For data fusion weights, For compensation coefficient, For data convolution operators; the Including architectural style parameters and component combination parameters, This includes parameter correlation and parameter accuracy data. The range of values ​​is The range of values ​​is Adjustments are made dynamically based on data interaction latency and data volume.

[0037] Specifically, the data interaction mechanism between the Hektar architectural design gene bank driving platform and the architectural model family parameter generation module is clearly defined. By setting reasonable data fusion weights, compensation coefficients, and convolution operator application methods, efficient and accurate data transfer between the two is achieved, avoiding information loss or data format incompatibility issues during data interaction. Simultaneously, compensation coefficients are used to dynamically adjust data deviations, ensuring that the interacted data accurately reflects the historical design experience in the gene bank and the current characteristics of the intermediate parameter set. This provides comprehensive data support for the generation of target parameters for the architectural model family, improving the matching degree between target parameters and historical high-quality design schemes. During implementation, the gene bank data includes architectural style parameters (such as component proportion parameters corresponding to modern minimalist style, and decorative detail parameters corresponding to new Chinese style), and component combination parameters (such as connection methods between walls and floor slabs, and matching parameters between beams, columns, and walls). These data must first be formatted uniformly to ensure consistency with the format of the intermediate parameter set. The intermediate parameter set includes parameter correlation (values ​​ranging from 0.6 to 0.9, with higher values ​​indicating stronger parameter synergy) and parameter accuracy data (error range ±0.5% to ±2%). The data fusion weight is set to 0.3-0.7, adjusted according to data importance. Gene pool data accounts for 40%-60%, intermediate parameter sets account for 40%-60%, and the gene pool data proportion for complex building model families is increased to 55%-60%. The compensation coefficient is set to 0.1-0.4. When the data interaction delay exceeds 500 milliseconds or the data volume exceeds 1000 records, the compensation coefficient is increased to 0.3-0.4; when the delay is less than 200 milliseconds and the data volume is less than 500 records, it is decreased to 0.1-0.2. The data convolution operator is used for feature extraction and fusion of the two types of data. The fused data needs to undergo integrity verification to ensure that key parameters are not missing and the missing rate is less than 2% before it can be transmitted to the parameter generation module for subsequent mapping processing.

[0038] Preferably, the parameter mapping expression of the building model family parameter generation module is: ,in, Indicates the target parameters after mapping. For the intermediate parameter set, For a set of parameter mapping rules, For a set of mapping vectors, For the number of mapping rules, For the first The weight of each mapping rule, For the first Mapping rule function, For the first The mapping vector corresponding to each rule; This includes parameter priority and parameter compatibility data. Used to convert intermediate parameters into component parameters and attribute parameters for building model families. The value range is 0.2-0.8. for Dimensional vector, For the target parameter dimension.

[0039] Specifically, the parameter mapping function of the building model family parameter generation module clearly defines the weight allocation of mapping rules, the dimension setting of mapping vectors, and the numerical adjustment standards of the mapping process. This ensures that the module can accurately convert the intermediate parameter set into target parameters that the building model family can recognize, avoiding parameter failures caused by fuzzy mapping rules or mismatched vector dimensions. Simultaneously, reasonable weight settings highlight the mapping priority of key parameters, ensuring that the target parameters fully reflect the component attributes and assembly requirements of the building model family, supporting the direct application of the building model family. During implementation, the parameter priorities in the intermediate parameter set are determined based on their impact on the functionality of the building model family. For example, calibration parameters such as component load-bearing parameters have a priority weight of 0.7-0.8, while secondary parameters such as surface decoration parameters have a priority weight of 0.2-0.3. Parameter compatibility data is obtained by comparing the matching degree between intermediate parameters and target model parameter formats. Parameters with a matching degree higher than 90% are directly mapped, while those with a matching degree lower than 70% require format conversion first. The number of mapping rules is determined according to the parameter category: 15-20 rules for basic component parameters, 20-25 rules for geometric feature parameters, and 18-22 rules for assembly-related parameters. The weight of each mapping rule is adjusted based on the number of parameters it covers. Rules covering more parameters have a weight of 0.6-0.8, while those covering fewer parameters have a weight of 0.2-0.4. The mapping vector dimension is consistent with the target parameter dimension. If the target parameters include 8 dimensions such as component number, size, and material, the mapping vector is set to a 1x8 dimension. During the mapping process, if the value exceeds the target range, fine-tuning is required, with the adjustment range controlled within ±1%. For example, if the wall length of 8200 in the intermediate parameter exceeds the target range of 3000-8000, it needs to be adjusted to 8000 to ensure that all target parameters meet the application requirements of the building model family.

[0040] Preferably, the collaborative expression of the building boundary representation structured potential geometric tree diffusion model and the building code embedded parameter generation and verification hybrid algorithm is as follows: ,in, This indicates the output of collaborative work. For the output of the diffusion model, For the output of the verification algorithm, This is the correction value for the cooperative error. For collaborative weights; the Including diffusion parameter deviations and diffusion efficiency data, This includes data on verification accuracy and parameter adjustment suggestions. The range of values ​​is Dynamic calculation based on data deviations during the collaboration process.

[0041] Specifically, the collaborative working mechanism of the structured potential geometric tree diffusion model representing building boundaries and the hybrid algorithm for parameter generation and verification embedded in building codes achieves efficient fusion of the computational results by setting collaborative weights and error correction values. This avoids parameter contradictions or insufficient collaboration caused by the independent operation of the model and algorithm. Simultaneously, by dynamically calculating error correction values ​​to compensate for data deviations during the collaboration process, it ensures that the collaborative output reflects both the advantages of the geometric feature parameters generated by the model and the value of the algorithm's code verification. This provides a unified and accurate data basis for subsequent parameter optimization and secondary diffusion, improving the coherence and reliability of the entire parameter generation process. During implementation, the diffusion model output includes diffusion parameter deviations (within ±2% error range) and diffusion efficiency data (single-level diffusion time 2-5 seconds). These data need to be standardized first, converting the deviation values ​​into standardized values ​​between 0 and 1, and assigning values ​​of 0.1-0.9 to the diffusion efficiency data, sorted by time duration. The validation algorithm output includes validation accuracy (overall accuracy above 95%) and parameter adjustment suggestions (clearly defining the direction and range of parameter adjustments for each suggestion). These are also standardized: the accuracy is converted to a percentage value of 0-1, and the adjustment suggestions are assigned values ​​of 0.3-0.8 based on feasibility. The collaborative weight is set to 0.4-0.6, maintaining a balanced weighting between the model output and the algorithm output, typically each set to 0.5. The error correction value is dynamically calculated based on data deviations during the collaboration process. When the deviation rate between the model and algorithm outputs is below 5%, the correction value is set to 0.05-0.1; when the deviation rate is between 5% and 10%, the correction value is set to 0.1-0.2; when the deviation rate is above 10%, the model and algorithm's computation process needs to be re-examined, corrected, and the correction value recalculated. Ultimately, this ensures that the error rate of the collaborative output is controlled within 3%, meeting the parameter application requirements of subsequent steps.

[0042] Preferably, step S3 specifically includes the following sub-steps: Step S31: Retrieve a preset building code system from the storage module of the hybrid algorithm for generating and verifying embedded parameters in building codes. This system includes building structural safety codes, building fire protection codes, and building space use codes. Decompose each code into specific parameter constraints to form a list of parameter constraint entries; Step S32: Classify the geometric feature parameter subsets generated in step S2. Divide the parameter subsets into wall parameter subsets, floor slab parameter subsets, and beam and column parameter subsets according to the building component type. Label each subset with the corresponding component identification information; Step S33: Associate the classified parameter subsets with the list of parameter constraint entries formed in step S31. Compare each parameter in the parameter subset with the requirements of the corresponding constraint entry through the rule matching module, and record the matching results of the parameters and constraint entries; Step S34: Mark the parameters that do not conform to the constraint entries according to the matching results. Add a non-conformity reason label to the marked parameters. The label includes the corresponding constraint entry number and the specific numerical information of the parameter exceeding the range.

[0043] Specifically, step S3 ensures the standardization and accuracy of the standard verification process by decomposing specifications, classification parameters, matching rules, and marking non-compliant parameters. This avoids verification omissions or misjudgments caused by unclear operation procedures, providing a clear basis for parameter adjustment in S4 and ensuring that subsequent parameter generation meets the requirements of various building industry standards. Implementation strictly follows the four sub-steps of S3: S31 retrieves the preset standard system from the building standard embedded parameter generation and verification hybrid algorithm storage module. This system includes three categories of standards: structural safety, fire protection, and space use. It is decomposed into 280 parameter constraint items, forming an item list. The structural safety standard includes component load-bearing strength limits (e.g., the compressive strength of concrete components is not less than 20), and the fire protection standard includes the standard for the spacing of evacuation passage walls (not less than 1200). S32 classifies the three-level geometric feature parameter subsets generated in S2 according to component type, dividing them into wall, S33 sets up subsets of floor slab and beam / column parameters, each subset marked with a unique component identifier, such as walls marked "W-01 to W-50"; S33 associates the categorized parameter subsets with the constraint item list, comparing them one by one through the rule matching module, such as comparing wall spacing parameters with fire protection code items, and recording the matching results; S34 marks non-compliant parameters, adding tags including constraint item number (such as "XF-03") and the specific value of the parameter exceeding the range (such as "actual 900, limit 1200, deviation -300"), ensuring that each violation information is traceable and supporting the precise adjustment of S4.

[0044] Preferably, step S4 specifically includes the following sub-steps: Step S41: Extract historical parameter adjustment data of the same type as the marked parameters in step S3 from the historical database of the Hektar architectural design gene bank driven platform. This data includes the initial value of the historical parameters, the adjusted values, and the standard clauses on which the adjustment is based; Step S42: Input the extracted historical adjustment data into the parameter optimization module of the building code embedded parameter generation and verification hybrid algorithm. Analyze the patterns of the historical adjustment data through the data mining unit in the module to generate a parameter adjustment trend curve; Step S43: Based on the adjustment trend curve and the non-compliance reason labels of the marked parameters in step S3, construct a parameter adjustment objective function. This function aims to ensure that the parameters meet the standard requirements and are close to the historical optimal adjustment value; Step S44: Calculate the parameter adjustment range and direction based on the parameter adjustment objective function to form a specific parameter adjustment scheme. The scheme clearly defines the values ​​of each marked parameter before and after adjustment.

[0045] Specifically, step S4 standardizes the parameter adjustment process by extracting historical data, analyzing patterns, constructing objective functions, and generating schemes. This avoids the subjectivity of traditional adjustments that rely on experience, ensuring that the adjusted parameters not only comply with regulations but also coordinate with the compliant parameters in S3, laying a high-quality foundation for parameter integration in S5. Implementation follows the four steps of S4: S41 extracts historical adjustment data of similar parameters marked in S3 from the historical database of the Hektar architectural design gene bank-driven platform. The scope is limited to records of similar building model families from the past 5 years, extracting 150-200 valid data entries. Each entry includes the initial value of the historical non-compliant parameter (e.g., "beam-column width 280"), the adjusted value (e.g., "300"), and the regulatory clause on which the adjustment was based (e.g., "Structural Safety Code - JGJ-08"); S42 inputs the data into the algorithm parameter optimization module, analyzing patterns through the data mining unit, such as common wall length parameters. The adjustment range is 3%-8% of the original value, and the adjustment direction is mostly to increase, generating an adjustment trend curve; S43 constructs a function based on the curve and the non-compliance reasons label of the parameters marked in S3 (such as "wall length 6800 < standard limit 7000"), with the goal of "parameters complying with the standard, deviation from adjacent compliant parameters less than 3%, and close to the historical best value"; S44 calculates the adjustment range (such as "6800 needs to be increased to 7000, range +200") and direction based on the function, forming a scheme including the values ​​before and after the adjustment, ensuring that the adjustment of each marked parameter has a clear basis, and the adjusted parameters can be directly used for integration in S5.

[0046] Preferably, step S5 specifically includes the following sub-steps: Step S51: Collect the parameters corrected in step S4 and the parameters that conform to the rules in step S3, and perform format unification processing on the two types of parameters, adjusting the numerical format and unit expression (if any) of the parameters to a consistent standard to form a preliminary parameter set; Step S52: Perform parameter correlation analysis on the preliminary parameter set, calculate the correlation coefficient between parameters through the correlation analysis unit in the building boundary representation structured potential geometric tree diffusion model, and identify parameter groups with strong correlation; Step S53: Input the preliminary parameter set into the building boundary representation structured potential geometric tree diffusion model, set the number of levels and diffusion coefficients for secondary diffusion, start the model to perform secondary diffusion processing, and strengthen the collaborative diffusion of parameter groups with strong correlation during the diffusion process; Step S54: Integrate the parameters after secondary diffusion processing, remove redundant parameters generated during the diffusion process, retain the calibrated valid parameters, and form an intermediate parameter set for the building model family.

[0047] Specifically, step S5 ensures the completeness and coordination of the intermediate parameter set generated by S5 through format unification, correlation analysis, secondary diffusion, and redundancy removal. This strengthens the geometric correlation between parameters and prevents parameter format confusion or insufficient correlation from affecting parameter mapping in S6, providing reliable intermediate data for generating target parameters in S6. During implementation, the four sub-steps of S5 are strictly followed: S51 collects the corrected parameters from S4 (approximately 35-50) and compliant parameters from S3 (approximately 180-220), performs format unification processing, and standardizes the numerical format to two decimal places (e.g., "2400→2400.00", "350→350.00"), forming a preliminary parameter set (approximately 215-270 parameters); S52 calculates the correlation coefficient between parameters using the correlation analysis unit of the structured potential geometric tree diffusion model representing building boundaries, identifying parameter groups with correlation coefficients higher than 0.7 (e.g., "wall length and floor width"). (Beam and column height and floor slab thickness); S53 inputs the initial set into the model, sets the number of secondary diffusion layers to 2 and the diffusion coefficient to 0.4-0.6, and starts the model to strengthen the collaborative diffusion of the associated parameter group. For example, when the wall length increases by 200, the floor slab width increases by 150 simultaneously; S54 integrates the secondary diffusion parameters, removes redundant parameters (such as duplicate "wall thickness-01" and "wall thickness-02"), retains the calibrated valid parameters, and forms an intermediate parameter set of the building model family (about 200-250 parameters), ensuring that the parameters are not redundant and have strong geometric correlation, which can be directly used for the mapping processing in S6.

[0048] The building boundary representation structured latent geometric tree diffusion model in this invention is an AI-driven model for processing the geometric features of building model family parameters. It can transform basic calibration parameters into a multi-dimensional, hierarchical subset of geometric feature parameters, thereby enhancing the geometric correlation of parameters. The implementation process requires multiple steps: First, the calibration parameters filtered by S1 (such as component length 2400-7200, width 3000-6000) are received, and a multi-dimensional geometric feature diffusion network of 1-10 layers is constructed. The layer is selected according to the complexity of the component (3-5 layers for simple components, 8-10 layers for complex components). Then, the diffusion coefficient (value 0.1-0.8, decreasing from 0.7-0.8 to 0.1-0.2 as the layer increases) and the geometric feature layer weights are set (overall contour layer accounts for 40%-50%, local detail layer accounts for 30%-40%, and interface connection layer accounts for 10%-20%). Layered diffusion is completed through activation functions and weight matrices (50-80 dimension matching calibration parameters and 30-50 output parameters). Finally, in S5, the integrated parameters are diffused a second time (2 layers, coefficient 0.4-0.6) to strengthen parameter correlation (parameter groups with correlation coefficients higher than 0.7 are adjusted collaboratively). The purpose of this model is to generate accurate geometric feature parameters, providing detailed evidence for standard verification while avoiding the problem of weak geometric correlation between parameters. Its significance lies in overcoming the limitation of single geometric feature representation in traditional parameter processing, improving the completeness and accuracy of the geometric description of building model family parameters, and supporting the reliability of subsequent target parameter generation.

[0049] The building code-embedded parameter generation and verification hybrid algorithm in this invention is a verification algorithm that integrates building codes and parameter processing. It can simultaneously complete compliance verification and adjust non-compliant parameters during parameter generation. Its implementation requires phased implementation: In S3, a verification rule library containing 280 rules is first retrieved (structural safety codes such as concrete compressive strength not less than 20, fire protection codes such as evacuation route spacing not less than 1200, and space use codes such as bedroom floor slab ratio not exceeding 70%). Following the order of "overall first, then local", the subset of geometric feature parameters generated in S2 (such as component position X-axis 0-10000, Y-axis 0-8000, angle 89.5°-90.5°) is matched with the rules, and violations are marked. The parameters (labeling rate 8%-12%) are marked and deviation labels are added (e.g., "Actual 900, Limit 1200, Deviation -300"). In S4, 150-200 historical adjustment data points from the past 5 years are extracted. Through data mining analysis of patterns (e.g., wall length adjustment range 3%-8%), a function is constructed with the objective of "compliance with specifications and deviation from compliant parameters less than 3%" to calculate adjustment schemes (e.g., "6800→7000, range +200"). The secondary verification pass rate must reach 100%. The purpose of this algorithm is to ensure that parameters comply with industry standards, reduce subsequent adjustment costs, and provide data support for parameter optimization. It solves the problems of fragmented application of standards and reliance on experience in traditional verification, improves the systematic nature and accuracy of parameter verification, and ensures the compliance and security of building model family parameters.

[0050] The Hektar architectural design gene bank driving platform in this invention is a support platform that stores and calls up historical architectural design data and gene information, providing a data foundation for parameter extraction, adjustment and interaction. Its implementation requires multi-stage collaboration: In S1, the data extraction function is initiated, a range is set (e.g., a high-rise residential building model family), and basic parameters (geometric features, material properties, spatial layout parameters) are extracted. A filtering module retains calibration parameters with a matching degree higher than 90% (60% geometric, 25% material, and 15% spatial), with a missing rate of less than 2%. In S4, 150-200 records of similar parameter adjustments from the past 5 years are retrieved from the historical database, providing initial values, adjusted values, and regulatory basis data. In the data interaction between S4 and S6, a fusion weight of 0.3-0.7 is set (55%-60% of the complex model family gene pool data) and a compensation coefficient of 0.1-0.4 (adjusted to 0.3-0.4 when the delay exceeds 500 milliseconds or the data volume exceeds 1000 records). Data fusion is completed through convolution operators, with a missing rate of less than 2% for integrity verification. The platform's role is to provide high-quality data sources, support parameter filtering, adjustment, and interaction, and ensure that parameter generation has historical experience for reference. By integrating scattered design data into standardized resources, the problem of insufficient data support in traditional parameter generation is solved, improving the efficiency and matching degree of parameter generation, and providing stable data support for AI-driven architectural model family parameter generation.

[0051] like Figure 2As shown, the AI-driven architectural model family parameter generation system includes: a Hektar architectural design gene bank parameter extraction unit, which is connected to the architectural boundary representation structured latent geometric tree diffusion model unit, used to extract the basic parameter set of the architectural model family from the Hektar architectural design gene bank driven platform and perform feature filtering, and transmit the filtered calibration parameters to the architectural boundary representation structured latent geometric tree diffusion model unit; an architectural boundary representation structured latent geometric tree diffusion model unit, which is connected to the Hektar architectural design gene bank parameter extraction unit and the building code embedded parameter verification unit, used to perform hierarchical diffusion processing on the calibration parameters to generate a subset of geometric feature parameters, and transmit this subset to the building code embedded parameter verification unit; and a building code embedded parameter verification unit, which is connected to the architectural boundary representation structured latent geometric tree diffusion model unit and the parameter optimization and adjustment unit, used to... The system verifies a subset of geometric feature parameters according to building codes and marks parameters that do not conform to the rules. The marked parameters are then transmitted to the parameter optimization and adjustment unit. This unit, connected to both the building code-embedded parameter verification unit and the parameter integration secondary diffusion unit, generates parameter adjustment schemes based on historical data and corrects the marked parameters. The corrected parameters are then transmitted to the parameter integration secondary diffusion unit. The parameter integration secondary diffusion unit, connected to both the parameter optimization and adjustment unit and the building model family target parameter generation unit, integrates the conforming parameters and the corrected parameters. It then performs secondary diffusion on the building boundary representation structured potential geometric tree diffusion model, generating an intermediate parameter set, which is transmitted to the building model family target parameter generation unit. Finally, the building model family target parameter generation unit, connected to the parameter integration secondary diffusion unit, maps the intermediate parameter set to generate target parameters recognizable by the building model family.

[0052] A method and system for generating architectural model family parameters based on AI has been developed. The primary advantage of this method and system lies in its construction of a parameter extraction and reuse mechanism calibrated by an architectural design gene bank-driven platform. This effectively overcomes the shortcomings of existing technologies that lack in-depth utilization of design gene data. By calling this platform to extract basic parameters of the architectural model family and filter and calibrate parameters, it integrates scattered historical design data into standardized resources. This allows the parameter generation process to fully leverage historical design experience, forming a reusable parameter generation logic. This process eliminates the need for subjective human judgment in parameter selection, reducing efficiency differences in parameter generation between different projects and improving the matching degree between generated parameters and historical high-quality design schemes. It solves the problems of low efficiency and susceptibility to subjective factors in traditional manual parameter generation, providing a stable and efficient data source for parameter generation.

[0053] Another major advantage of this method and system is the deep integration of building codes and parameter generation, as well as the enhanced geometric correlation of parameters. This overcomes the shortcomings of existing technologies, such as lagging code verification and easily compromised parameter correlation. It integrates a hybrid algorithm for parameter generation and verification embedded in building codes into the entire parameter generation process, performing code verification synchronously during parameter generation, rather than independently after generation. This allows for direct marking and adjustment of parameters that do not conform to the codes, reducing repeated corrections and lowering process complexity. Simultaneously, by leveraging a structured potential geometric tree diffusion model representing building boundaries, it performs hierarchical and secondary diffusion processing on parameters, strengthening the geometric correlation between parameters and preventing later adjustments from disrupting the synergistic relationship between parameters. This ensures the overall coordination and accuracy of the building model family, significantly improving the precision of parameter generation and meeting the architectural industry's requirements for the standardization and completeness of design schemes.

[0054] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," "link," and "fix" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0055] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for generating architectural model family parameters based on AI, characterized in that, include: Step S1: Call the Hektar architectural design gene library driving platform to extract the basic parameter set of the architectural model family. This set includes geometric feature parameters of building components, material property parameters, and spatial layout related parameters. The extracted basic parameter set is filtered by the platform's built-in data filtering module, retaining the calibration parameters directly related to the generation of the architectural model family. Step S2: Input the calibration parameters filtered in Step S1 into the architectural boundary representation structured potential geometric tree diffusion model to construct a multi-dimensional geometric feature diffusion network. In this network, the geometric features corresponding to the calibration parameters are subjected to hierarchical diffusion processing to generate geometric feature parameter subsets at different diffusion levels. Step S3: Input the geometric feature parameter subsets generated in Step S2 into the building code embedded parameter generation and verification hybrid algorithm. This algorithm first constructs a parameter verification rule base based on the preset building code system, and then uses a rule matching module to match the geometric feature parameter subsets with the verification rule base. The rules are matched one by one, and parameters that do not conform to the rules are marked. Step S4: Adjust the parameters that do not conform to the rules marked in step S3. Based on the parameter optimization module in the hybrid algorithm for parameter generation and verification embedded in the building code, and combined with the historical optimization data stored in the Hektar building design gene bank driving platform, a parameter adjustment scheme is generated, and the parameters that do not conform to the rules are corrected according to the scheme. Step S5: Integrate the parameters corrected in step S4 with the parameters that conform to the rules to form the initial parameter set of the building model family. Input the initial parameter set into the building boundary representation structured potential geometric tree diffusion model for secondary diffusion processing to strengthen the geometric correlation between parameters and generate the intermediate parameter set of the building model family. Step S6: Perform parameter mapping processing on the intermediate parameter set generated in step S5. Through the building model family parameter generation module, the intermediate parameter set is mapped to the target parameters that the building model family can recognize, and the building model family parameter generation is completed.

2. The method for generating building model family parameters based on AI according to claim 1, characterized in that, The expression representing the structured potential geometric tree diffusion model for the building boundary is: ,in, This represents the post-diffusion geometric feature parameters output by the model. These are the calibration parameters selected in step S1. The number of diffusion levels. The diffusion coefficient of the model is . The number of layers for geometric features, For the first Weighting coefficients of layer geometric features For the first Layer geometric feature diffusion function, For activation function, For the first The weight matrix of the layer; Including building component length parameters Width parameters and height parameters The value range is 1-10. The value range is 0.1-0.

8. satisfy for Dimensional matrix, To calibrate the number of parameters, For the first Number of layer output parameters.

3. The method for generating building model family parameters based on AI according to claim 1, characterized in that, The expression for the hybrid algorithm for generating and validating embedded parameters in the building code is as follows: ,in, This represents the parameter verification result output by the algorithm. The subset of geometric feature parameters generated in step S2, For the set of rules in the parameter validation rule base, This is a historical data set within the Hektar architectural design gene bank-driven platform. To verify the number of rules, For the first The validation weight of each rule For the first Validation rule function, For data fusion operators, In order to be with the first The historical data corresponding to the rule; Including component position parameters and angle parameters , Used for judgment Does it comply with building spacing specifications and component size specifications? The range of values ​​is This includes the pass rate and adjustment range data for similar parameters in the past.

4. The method for generating building model family parameters based on AI according to claim 1, characterized in that, The data interaction expression between the Hektar architectural design gene bank driving platform and the architectural model family parameter generation module is as follows: ,in, This represents the data after the interaction. For gene bank data in the platform, The intermediate parameter set generated in step S5 For data interaction compensation coefficient, For data fusion weights, For compensation coefficient, For data convolution operators; the Including architectural style parameters and component combination parameters, This includes parameter correlation and parameter accuracy data. The range of values ​​is The range of values ​​is Adjustments are made dynamically based on data interaction latency and data volume.

5. The method for generating building model family parameters based on AI according to claim 1, characterized in that, The parameter mapping expression of the building model family parameter generation module is as follows: ,in, Indicates the target parameters after mapping. For the intermediate parameter set, For a set of parameter mapping rules, For a set of mapping vectors, For the number of mapping rules, For the first The weight of each mapping rule, For the first Strip mapping rule function, For the first The mapping vector corresponding to each rule; This includes parameter priority and parameter compatibility data. Used to convert intermediate parameters into component parameters and attribute parameters for building model families. The value range is 0.2-0.

8. for Dimensional vector, For the target parameter dimension.

6. The method for generating building model family parameters based on AI according to claim 1, characterized in that, The collaborative expression of the building boundary representation structured potential geometric tree diffusion model and the building code embedded parameter generation and verification hybrid algorithm is as follows: ,in, Indicates the output of collaborative work. For the output of the diffusion model, For the output of the verification algorithm, This is the correction value for the cooperative error. For collaborative weights; the Including diffusion parameter deviations and diffusion efficiency data, This includes data on verification accuracy and parameter adjustment suggestions. The range of values ​​is Dynamic calculation based on data deviations during the collaboration process.

7. The method for generating building model family parameters based on AI according to claim 1, characterized in that, Step S3 specifically includes the following sub-steps: Step S31: Retrieve the preset building code system from the storage module of the hybrid algorithm for generating and verifying embedded parameters in building codes. This system includes building structural safety codes, building fire protection codes, and building space use codes. Decompose each code into specific parameter constraints to form a list of parameter constraint entries; Step S32: Classify the geometric feature parameter subsets generated in Step S2. Divide the parameter subsets into wall parameter subsets, floor slab parameter subsets, and beam and column parameter subsets according to the building component type. Label each subset with the corresponding component identification information; Step S33: Associate the classified parameter subsets with the list of parameter constraint entries formed in Step S31. Compare each parameter in the parameter subset with the requirements of the corresponding constraint entry through the rule matching module, and record the matching results of the parameters and constraint entries; Step S34: Mark the parameters that do not conform to the constraint entries according to the matching results. Add a non-conformity reason label to the marked parameters. The label includes the corresponding constraint entry number and the specific numerical information of the parameter exceeding the range.

8. The method for generating building model family parameters based on AI according to claim 1, characterized in that, Step S4 specifically includes the following sub-steps: Step S41: Extract historical parameter adjustment data of the same type as the marked parameters in Step S3 from the historical database of the Hektar Architectural Design Gene Bank Driven Platform. This data includes the initial value of the historical parameters, the adjusted values, and the standard clauses on which the adjustment is based. Step S42: Input the extracted historical adjustment data into the parameter optimization module of the building code embedded parameter generation and verification hybrid algorithm. Analyze the patterns of the historical adjustment data through the data mining unit in the module to generate a parameter adjustment trend curve. Step S43: Based on the adjustment trend curve and the non-compliance reason labels of the marked parameters in Step S3, construct a parameter adjustment objective function. This function aims to ensure that the parameters meet the standard requirements and are close to the historical optimal adjustment value. Step S44: Calculate the parameter adjustment range and direction based on the parameter adjustment objective function to form a specific parameter adjustment plan. The plan specifies the values ​​of each marked parameter before and after adjustment.

9. The method for generating building model family parameters based on AI according to claim 1, characterized in that, Step S5 specifically includes the following sub-steps: Step S51: Collect the parameters corrected in Step S4 and the parameters that conform to the rules in Step S3. Standardize the format of both types of parameters, adjusting the numerical format and unit representation (if any) to a consistent standard to form a preliminary parameter set. Step S52: Perform parameter correlation analysis on the preliminary parameter set. Calculate the correlation coefficient between parameters using the correlation analysis unit in the building boundary-represented structured potential geometric tree diffusion model, identifying parameter groups with strong correlations. Step S53: Input the preliminary parameter set into the building boundary-represented structured potential geometric tree diffusion model, set the number of levels and diffusion coefficients for secondary diffusion, and start the model for secondary diffusion processing. Strengthen the collaborative diffusion of parameter groups with strong correlations during the diffusion process. Step S54: Integrate the parameters after secondary diffusion processing, removing redundant parameters generated during the diffusion process and retaining calibrated valid parameters to form an intermediate parameter set for the building model family.

10. An AI-driven architectural model family parameter generation system, characterized in that, include: The Hektar Architectural Design Gene Bank Parameter Extraction Unit, connected to the Architectural Boundary Representation Structured Potential Geometric Tree Diffusion Model Unit, is used to extract the basic parameter set of the architectural model family from the Hektar Architectural Design Gene Bank driving platform and perform feature filtering. The filtered calibration parameters are then transmitted to the Architectural Boundary Representation Structured Potential Geometric Tree Diffusion Model Unit. The Architectural Boundary Representation Structured Potential Geometric Tree Diffusion Model Unit, connected to both the Hektar Architectural Design Gene Bank Parameter Extraction Unit and the Building Code Embedded Parameter Verification Unit, performs hierarchical diffusion processing on the calibration parameters to generate a subset of geometric feature parameters, which is then transmitted to the Building Code Embedded Parameter Verification Unit. The Building Code Embedded Parameter Verification Unit, connected to both the Architectural Boundary Representation Structured Potential Geometric Tree Diffusion Model Unit and the Parameter Optimization and Adjustment Unit, verifies the subset of geometric feature parameters according to building codes and marks parameters that do not conform to the rules, transmitting the marked parameters to the Parameter Optimization and Adjustment Unit. The parameter optimization and adjustment unit is connected to the building code embedded parameter verification unit and the parameter integration secondary diffusion unit. It is used to generate parameter adjustment schemes based on historical data and correct the marked parameters, and then transmit the corrected parameters to the parameter integration secondary diffusion unit. The parameter integration secondary diffusion unit is connected to the parameter optimization and adjustment unit and the building model family target parameter generation unit, respectively. It is used to integrate the rule-compliant parameters and the correction parameters, input the building boundary to represent the structured potential geometric tree diffusion model for secondary diffusion, generate an intermediate parameter set and transmit it to the building model family target parameter generation unit. The target parameter generation unit for the building model family is connected to the parameter integration secondary diffusion unit. It is used to perform parameter mapping on the intermediate parameter set and generate target parameters that can be recognized by the building model family.

Images