Enterprise knowledge base data-driven residential entrance facade generation method and interactive terminal

Abstract

The present application relates to the technical field of computer-aided design, in particular to a residential entrance facade generation method driven by enterprise knowledge base data and an interactive terminal, comprising the following steps: extracting facade key parameter classification and collection, setting threshold to screen samples to generate a training set, coding parameter card to identify difference sorting, and extracting path ratio to generate a parameter combination set; in the present application, through standardized extraction and attribute classification of residential entrance facade project parameters, structure and component feature collection is realized, sample comparability is improved, representative training samples are screened by setting construction threshold and proportion rules, parameter learning effect is strengthened, efficient index structure is constructed through field coding and parameter set, data retrieval and combination efficiency is improved, logical connection between details and overall structure is enhanced, difference sorting and offset ratio analysis mechanism realize rapid fitting and matching of input parameters, shorten the conception cycle, and improve design response speed, result controllability and scheme reuse value.

Description

Technical Field

[0001] This invention relates to the field of computer-aided design technology, and in particular to a method and interactive terminal for generating residential entrance facades driven by enterprise knowledge base data. Background Technology

[0002] Computer-aided design (CAD) technology refers to the technical activities of using computer hardware and software to perform geometric modeling, graphic generation and modification, material rendering, and parametric expression of objects such as buildings, industries, and engineering projects. Its core aspects include visually constructing design schemes through computer programs, combining and transforming design elements using data-driven methods, and providing reusable knowledge and parameter support at different design stages. This technical field typically employs methods such as 3D modeling, image generation, parametric modeling, and rendering engines to assist the design process. Traditional residential entrance facade design, in particular, relies primarily on manual modeling tools and general image generation tools during the conceptualization and expression of building facades, typically generated using 3D modeling software. While structural frameworks can be used to achieve material and lighting effects with the help of rendering tools, or to quickly generate images using general AI drawing systems, these methods often lack architectural expertise and suffer from improper proportion control. Furthermore, while architectural rendering tools can provide realistic visual effects, they rely on manual, step-by-step modeling, a complex process that makes it difficult to accumulate design knowledge. Vertical design generation tools primarily generate building plans and elevations through algorithms, but have not yet formed a complete closed-loop process encompassing data acquisition, image generation, and parameter optimization. This invention also relates to the field of AI multimodal collaborative design, integrating visual recognition, semantic segmentation, LoRA lightweight training, and a knowledge base data-driven mechanism to achieve intelligent and automated building facade generation.

[0003] Traditional residential entrance facade design relies heavily on manual modeling software and general image generation tools. The lack of parameter support specific to architectural structural logic during the graphic construction process leads to a high dependence on the designer's experience and subjective judgment, easily resulting in imbalances in scale and proportion. For example, when extending eaves or setting railings, failure to refer to scale parameters validated in previous projects can easily cause visual disharmony or structural node conflicts, affecting the overall harmony of the facade. Traditional design tools have fragmented workflows, requiring manual modeling, rendering, and modification, lacking a unified AI collaboration mechanism. General AI drawing models cannot understand architectural semantics, while vertical architectural AI platforms lack a closed loop of "collection → creation → refinement → knowledge accumulation." This invention bridges this gap through multimodal acquisition and AI intelligent collaboration. Furthermore, while existing tools can generate high-precision rendered images, the lack of systematic correlation between graphics and data prevents the structured accumulation and reuse of parameters, causing knowledge gaps in the design process and making it difficult to transfer existing experience across multiple projects. While general-purpose intelligent drafting tools possess image generation capabilities, they lack the ability to understand specialized architectural parameters. The generated results often have deficiencies in terms of the actual structural feasibility of buildings, failing to meet the needs of detailed design and construction drawing refinement. This approach restricts both design efficiency and accuracy, hindering the construction of a sustainable, iteratively optimized design system. General-purpose AI drafting tools (such as Midjourney) cannot understand the design logic of residential entrances, resulting in a high rate of scale imbalance. The tool workflow is fragmented: switching between multiple tools is required from data collection to generation to modification, leading to low efficiency. Furthermore, enterprise design parameters are difficult to accumulate, resulting in a knowledge reuse rate of less than 30%. Summary of the Invention

[0004] To address the technical problems existing in the prior art, embodiments of the present invention provide a method for generating residential entrance facades driven by enterprise knowledge base data.

[0005] This invention constructs a closed-loop system through a technical architecture of "enterprise knowledge base data-driven + AI multimodal collaboration," which forms a complete process from "inspiration collection – model training – intelligent generation – semantic refinement – ​​knowledge iteration," thereby improving the efficiency and consistency of building facade generation.

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

[0007] In a preferred embodiment of the present invention, by modeling and training data of benchmark residential projects of enterprises, a high-quality entrance facade drawing of the residential area is generated using a combination of StableDiffusion+LoRA. At the same time, multimodal data annotation and semantic segmentation technology are introduced, and combined with the parameter-driven method of the present invention, a two-way collaborative mechanism of "parameter optimization + intelligent generation" is formed.

[0008] This invention includes the following steps:

[0009] S1: Extract column spacing, eaves extension line length, window area ratio, railing height and material reflectivity of the residential entrance facade project from the enterprise knowledge base, unify the field format, classify them into structural parameter set, node construction set and detailed component set according to attributes, and generate a structured parameter collection table.

[0010] S2: Based on the column spacing, eaves extension line, window area ratio, railing height and doorway width in the structured parameter collection table, set the structural ratio and construction threshold conditions, add an identifier to the samples that meet any of the preset rules, and form a facade structure training sample set;

[0011] S3: Based on the training sample set of the facade structure, extract the window area ratio, material reflectivity and floor height parameters, divide them into sets of openings, materials and sizes according to parameter type, establish the correspondence between field codes and sample numbers, and collect the coded data into training sample parameter card set;

[0012] S4: Based on the set encoding data in the training sample parameter card set, identify the differences in the window area ratio, material reflectivity and floor height parameter sets, generate a sorting structure based on the differences between sets, establish a correspondence with the sample card index, and generate a fit ranking index number.

[0013] As a further aspect of the present invention, the structured parameter collection table includes column spacing parameters, eaves extension line length parameters, window area ratio parameters, railing height parameters, and material reflectivity parameters; the facade structure training sample set includes structural proportion rule identifiers, construction threshold condition identifiers, and sample number identifiers; the training sample parameter card set includes opening parameter set, material parameter set, and size parameter set; and the fit ranking index number includes difference ranking structure, parameter difference value, and sample card index.

[0014] As a further aspect of the present invention, the specific steps of S1 are as follows:

[0015] S101: Obtain the column spacing, eaves extension line length, window area ratio, railing height and railing material reflectivity parameters in the residential entrance facade project, unify the parameter format to standard numerical type, classify the fields according to attribute dimensions, establish the interval judgment rules for the corresponding fields, and generate standardized interval values ​​for parameter fields.

[0016] S102: Based on the column spacing and eaves extension line length in the normative interval values ​​of the parameter field, extract the structural line position parameters, divide the node segment numbers, summarize the railing height and material reflectivity under the corresponding segment, and establish the correspondence between railing parameters and node segments to obtain the railing parameter ratio value of node segment.

[0017] S103: Call the railing height, railing material reflectivity and window area ratio from the railing parameter ratio values ​​of the node segment, combine and classify them according to the segment number, calculate the ratio between railing height and window area ratio, filter parameter combinations based on the upper limit of the railing window area ratio, and merge and integrate the railing parameters, node position parameters and field specification values ​​according to the three attribute sets to generate a structured parameter collection table.

[0018] As a further aspect of the present invention, the specific steps of S2 are as follows:

[0019] S201: Obtain the column spacing, eaves extension line, window area ratio, railing height and door opening width parameters from the structured parameter collection table, establish field classification rules based on functional attributes, set threshold conditions related to structural proportion and construction, generate corresponding judgment criteria for fields, and obtain parameter condition judgment standard values.

[0020] S202: Call the threshold condition in the parameter condition judgment standard value, judge the parameter records in the structured parameter collection table one by one, mark the sample that meets the preset rule for any field, write the marking status into the field information, and obtain the sample quantity triggered by the construction ratio rule.

[0021] S203: Based on the constructed ratio rule, trigger the labeling information in the sample size, organize the labeled samples according to the field structure, remove unlabeled records, and classify the difference type fields into the corresponding sets, unify the sample set format, and generate a surface structure training sample set.

[0022] As a further aspect of the present invention, the specific steps of S3 are as follows:

[0023] S301: Based on the training sample set of the facade structure, extract the window area ratio, material reflectivity and floor height parameters, complete the parameter positioning according to the field order, merge the extracted fields into a unified set, and define the classification label according to the parameter attributes to obtain the parameter type belonging label value;

[0024] S302: Call the classification information of the field in the parameter type attribution mark value, divide the parameters into opening, material and size sets according to attribute characteristics, set classification numbering rules and complete the mapping, and obtain the parameter classification mapping number value;

[0025] S303: Based on the association structure between the field numbers contained in the parameter classification mapping number value and the sample number, establish the index relationship between the coded data and the sample record, organize the set field numbers and put them into a unified structure table, and generate the sample parameter card index value.

[0026] As a further aspect of the present invention, the specific steps of S4 are as follows:

[0027] S401: Based on the encoded data of the training sample parameter card set, call the window area ratio, material reflectivity and layer height parameter fields in the sample, match them with the corresponding fields in the input set, identify the numerical deviation between the fields, and obtain the parameter field difference value;

[0028] S402: Based on the difference values ​​of the parameter fields, identify the overall deviation of the corresponding fields of the samples, construct a sorting sequence according to the deviation, mark the corresponding positions between the sorting number and the sample number, and obtain the field difference sorting sequence;

[0029] S403: Based on the order of sample numbers in the sorted sequence of the field differences, construct a position mapping table in the index structure, verify the correspondence between the sorted numbers and the sample card numbers, and generate the fit ranking index number.

[0030] As a further aspect of the present invention, the method further includes:

[0031] S5: Based on the path results associated with the fitting degree sorting index number, extract the column spacing, window area ratio and railing height parameters in the corresponding path, determine the ratio offset with the current input parameters, and use the set of path parameters that meet the combination threshold conditions as the basis for structural combination to generate a set of parameter combinations for the entrance facade scheme of the residential area.

[0032] The set of parameters for the entrance facade design of the residential area includes column spacing ratio, window area ratio offset value, and railing height ratio;

[0033] The ratio offset determination refers to comparing the degree of offset between the ratios of two parameters and determining whether they are close based on a preset offset threshold;

[0034] The combined threshold condition refers to the maximum acceptable offset range of multiple parameter ratios in the structural combination, which is used for scheme screening and path fitting judgment.

[0035] The structural combination refers to the set of matched parameters used to generate the building facade design, including proportions, dimensions, and material elements.

[0036] As a further aspect of the present invention, the specific steps of S5 are as follows:

[0037] S501: Based on the path results associated with the sorted index number of the fit degree, extract the column spacing, window area ratio and railing height parameters corresponding to the path, match them with the corresponding fields in the input parameter set, identify the ratio relationship between each field, and obtain the path parameter ratio group;

[0038] S502: Based on the path parameter ratio group, and combined with the column spacing ratio threshold, window area ratio threshold, and railing height ratio threshold set in the input parameter set, determine whether the conditions of the path ratio combination are met, and obtain the threshold-satisfied path set;

[0039] S503: Based on the path number in the path set that meets the threshold, summarize the column spacing, window area ratio and railing height parameter fields under the corresponding path, combine them into the parameter set required for facade configuration, and generate the parameter combination set of the residential entrance facade scheme.

[0040] As a further aspect of the present invention, the window area ratio is the ratio of the window opening area in the building facade to the total area of ​​the facade in the same section;

[0041] The height of the railing is the vertical height from the tread surface to the top of the railing when railings are installed in open areas such as balconies and corridors, which conforms to the definition in the residential design code.

[0042] The reflectance of the material is the ratio of light reflected by the building facade material in the visible light band;

[0043] The structural ratio threshold condition refers to the ratio range established between structural parameters;

[0044] The construction threshold condition refers to the upper and lower bound standards set for the proportional or dimensional relationships between construction nodes;

[0045] The preset rules refer to the set of judgment conditions set according to the enterprise benchmark project, including the proportional threshold and size range parameter logic rules;

[0046] The field encoding refers to establishing a unique and identifiable encoding structure for each type of parameter in the parameter set through numbering, letters, or combinations thereof;

[0047] The difference value between sets refers to the number of content differences between the input parameter set and the coding set in the knowledge base, and is expressed by the set difference length or ratio.

[0048] The fit ranking index is a numbered sequence established from low to high parameter difference.

[0049] An intelligent interactive terminal for generating residential entrance facades, driven by enterprise knowledge base data and featuring AI multimodal collaboration, includes:

[0050] The structural parameter collection module obtains the column spacing, eaves extension line length, window area ratio, railing height and material reflectivity from the residential entrance facade project. The parameters are divided into structural parameter sets, node construction sets and detailed component sets according to their attributes. After unifying the field format, the parameters are merged and organized to generate a structured parameter collection table.

[0051] The structural rule filtering module, based on the column spacing, eaves extension line length, window area ratio, railing height and doorway width in the structured parameter collection table, sets structural proportions and construction threshold conditions, compares the relationship between parameters and thresholds, identifies sample records that satisfy any rule, and generates a floor structure training sample set.

[0052] The facade sample construction module calls the window area ratio, material reflectivity and floor height parameters in the facade structure training sample set, divides the three parameters into the opening set, material set and size set respectively, establishes the correspondence between field code and sample number, and organizes them into a training sample parameter card set.

[0053] The parameter difference fitting module uses the field encoding data in the training sample parameter card set to call the window area ratio, material reflectivity and layer height in the input parameters, calculates the field difference value and ratio, generates a sorting structure based on the difference value, establishes an index relationship between the sorting result and the sample card, and generates a good fit sorting index number.

[0054] The scheme combination generation module calls the column spacing, window area ratio and railing height in the path corresponding to the fitting degree sorting index number. Based on the ratio offset relationship between the current input parameters and the path parameters, it determines whether the combination meets the structural combination threshold conditions, filters the path parameter set and generates the residential entrance facade scheme parameter combination set.

[0055] Based on parameter-driven design, this invention further combines enterprise knowledge base and multimodal AI technology to propose an intelligent generation method for residential entrance facades based on the LoRA model, which can achieve high-quality image generation and semantic refinement within the existing parameter optimization framework.

[0056] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0057] In this invention, by standardizing the extraction and attribute classification of parameters for residential entrance facade projects, high-precision aggregation of structural and component features is achieved, improving the comparability between samples. By setting construction thresholds and proportion rules, representative training samples are accurately selected to enhance parameter learning effects. Through field coding and parameter sets, an efficient index structure is constructed to improve data retrieval and combination efficiency, enhance the logical connection between details and the overall structure, and the difference sorting and offset ratio analysis mechanism can achieve rapid fitting and matching of input parameters, significantly shortening the design cycle and improving design response speed, result controllability, and scheme reuse value. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a schematic diagram of the steps of the present invention;

[0060] Figure 2 This is a detailed schematic diagram of S1 of the present invention;

[0061] Figure 3 This is a detailed schematic diagram of S2 of the present invention;

[0062] Figure 4 This is a detailed schematic diagram of S3 of the present invention;

[0063] Figure 5 This is a detailed schematic diagram of S4 of the present invention;

[0064] Figure 6 This is a detailed schematic diagram of S5 of the present invention;

[0065] Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0066] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0067] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0068] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0069] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0070] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0071] Please see Figure 1 This invention provides a method for generating residential entrance facades driven by enterprise knowledge base data, comprising the following steps:

[0072] S1: Extract column spacing, eaves extension line length, window area ratio, railing height and material reflectivity of the residential entrance facade project from the enterprise knowledge base, unify the field format, classify them into structural parameter set, node construction set and detailed component set according to attributes, and generate a structured parameter collection table.

[0073] The window area ratio is the ratio of the window area in a building facade to the total area of ​​the same section of the facade. It is used to measure the relationship between the building's transparency and the facade composition.

[0074] The height of the railing is the vertical height from the tread surface to the top of the railing when railings are installed in open areas such as balconies and corridors, which complies with the definition in the "Residential Design Code".

[0075] Material reflectance is the ratio of light reflected by a building facade material in the visible light band, used to evaluate the visual brightness of the material;

[0076] S2: Based on the column spacing, eaves extension line, window area ratio, railing height and doorway width in the structured parameter collection table, set the structural ratio and construction threshold conditions, add labels to samples that meet any of the preset rules, and organize the labeled data into a facade structure training sample set.

[0077] Structural proportion threshold conditions refer to the ratio range established between structural parameters, used as a criterion to screen whether a building's structure conforms to a preset compositional proportion.

[0078] The threshold condition refers to the upper and lower bound standards set for the proportional or dimensional relationship between the constructed nodes, which are used to judge the rationality of the sample structure.

[0079] Preset rules refer to a set of judgment conditions based on parameters set according to the company's benchmark projects, including proportional thresholds and size range parameter logic rules;

[0080] S3: Extract window area ratio, material reflectivity and floor height parameters from the training sample set of facade structure, divide them into sets of opening, material and size according to parameter type, establish the correspondence between field code and sample number, and collect the coded data into training sample parameter card set;

[0081] Field encoding refers to establishing a unique and identifiable encoding structure for each type of parameter in a parameter set through numbering, letters, or combinations thereof;

[0082] S4: Based on the set encoding data in the training sample parameter card set, identify the differences in the input window area ratio, material reflectivity and layer height parameter sets, generate a sorting structure based on the differences between sets, establish an index correspondence with the sample cards, and generate a goodness-of-fit sorting index number.

[0083] The difference between sets refers to the amount of content difference between the set of input parameters and the set of codes in the knowledge base, and is expressed by the set difference length or ratio.

[0084] The fit ranking index is a numbered sequence established from low to high parameter difference, used to determine the matching priority of candidate parameter paths;

[0085] S5: Based on the path results associated with the index number sorted by fit, extract the column spacing, window area ratio and railing height parameters in the corresponding path, determine the ratio offset with the current input parameters, and use the set of path parameters that meet the combination threshold conditions as the basis for structural combination to generate the parameter combination set of the residential entrance facade scheme.

[0086] Ratio offset determination refers to comparing the degree of offset between the ratios of two parameters and determining whether they are close based on a preset offset threshold;

[0087] The combined threshold condition refers to the maximum acceptable offset range of the ratio of multiple parameters in a structural combination, which is used for scheme selection and path fitting judgment.

[0088] Structural composition refers to the set of matched parameters used to generate building facade designs, including proportions, dimensions, and material elements.

[0089] The structured parameter collection table includes column spacing parameters, eaves extension line length parameters, window area ratio parameters, railing height parameters, and material reflectivity parameters. The facade structure training sample set includes structural proportion rule identifiers, construction threshold condition identifiers, and sample number identifiers. The training sample parameter card set includes opening parameter set, material parameter set, and size parameter set. The fit ranking index number includes difference ranking structure, parameter difference value, and sample card index. The residential entrance facade scheme parameter combination set includes column spacing ratio, window area ratio offset value, and railing height ratio.

[0090] Please see Figure 2 The specific steps of S1 are as follows:

[0091] S101: Obtain the column spacing, eaves extension line length, window area ratio, railing height and railing material reflectivity parameters in the residential entrance facade project, unify the parameter format to standard numerical type, classify the fields according to attribute dimensions, establish the interval judgment rules for the corresponding fields, and generate standardized interval values ​​for parameter fields.

[0092] In the process of modeling the entrance facade of the residential area, the key parameters of the structure and components are first extracted, including column spacing, eaves extension line length, window area ratio, railing height, and the reflectivity of the railing material. During the operation, the facade is modeled using a BIM platform or drawings to obtain the geometric information of the components. For example, the column center spacing can be calculated by measuring each column individually. If the columns are arranged sequentially at 1m, 3.5m, 6.5m, and 9.5m along the X-axis, three sets of column spacings can be obtained, which are 2.5m, 3m, and 3m respectively. The eaves extension line length is determined by measuring the diagonal line length between the wall base and the eaves. If the bottom point is at a height of 0m, the top point is at a height of 3m, and the horizontal distance is 6m, then the extension line is approximately 6.7m. The window area ratio is determined by the ratio of the total window area to the wall area of ​​the same section. If the total window area is 8m² and the corresponding wall area is 20m², then the ratio is 40%. The railing height is directly determined by the elevation... The measurements are obtained from cross-sectional views. For example, common railing height values ​​include 1.05m, 1.1m, and 1.2m. The reflectivity of the railing material can be defined according to the standard material library, such as 0.6 for aluminum alloy and 0.8 for glass. After the parameters are uniformly converted into numerical formats, they are classified into structural categories (column spacing, extension line length), enclosure categories (window area ratio), and safety construction categories (railing height and material reflectivity) based on functional attributes. An interval value judgment standard is established for each type of field. For example, the column spacing is divided into segments such as 2.0 to 2.9m and 3.0 to 3.9m. The concentration of the sample is statistically analyzed. If the sample proportion in the 3.0 to 3.9m interval is 70%, then this interval can be set as the standard value. The railing height is set to the interval of 1.05m to 1.2m, and the railing reflectivity is divided into the interval of 0.5 to 0.8. The actual value of the component is compared with whether it is within the corresponding standard interval. If the column spacing is 3.5m, it is within the compliant interval. If it exceeds the standard, it is marked as deviating from the standard.

[0093] S102: Based on the column spacing and eaves extension line length in the parameter field normative interval value, extract the structural line position parameters, divide the node segment number, summarize the railing height and material reflectivity under the corresponding segment, and establish the correspondence between railing parameters and node segments to obtain the railing parameter ratio value of node segment.

[0094] Parameters are extracted from the structural lines of the nodes defined by the column spacing and the length of the eaves extension line. Based on these parameters, numbered sections are divided. Each section of component gap or geometric extension area is defined as a numbered node section. For example, with a facade length of 24m and a column spacing of 3m, 8 numbered sections can be divided, numbered N1 to N8. The railing height and the corresponding material reflectivity are recorded in each section. For example, N1 to N8 are set to railing heights ranging from 1.1m to 1.2m, and the materials are alternately set as aluminum alloy and glass. The ratio relationship between railing parameters and numbers is constructed in sequence. For example, N1 corresponds to 1.1m and 0.6 reflectivity, and N2 corresponds to 1. Using ratios such as 2m and 0.8 reflectivity, a matching set of railing parameters and structural zones is established. The ratio value is expressed as the ratio between railing height and reflectivity. For example, 1.1 divided by 0.6 yields approximately 1.83 as the parameter matching value. If the railing height is 1.2m and the material is glass, the ratio value is 1.5. The railing parameter ratio values ​​are recorded for each node segment, and an upper limit judgment standard is set. For example, the ratio value should not be higher than 2.0. If the ratio value is greater than this upper limit, the parameters of that segment are marked as abnormal. Under the entire structural line division, the correspondence between railing parameters and their numbers is extracted and summarized segment by segment, which can be used for subsequent parameter combination analysis.

[0095] S103: Call the railing height, railing material reflectivity and window area ratio from the railing parameter ratio values ​​of the node segment, combine and classify them according to the segment number, calculate the ratio between railing height and window area ratio, filter parameter combinations based on the upper limit of the railing window area ratio, and merge and integrate the railing parameters, node position parameters and field specification values ​​according to the three attribute sets to generate a structured parameter collection table;

[0096] The railing parameter ratios extract railing height, railing material reflectivity, and the corresponding window area ratio. These are then grouped and categorized according to pre-defined node numbers. Each numbered segment corresponds to a set of data, including railing height, reflectivity, and window area ratio. Dividing the railing height by the window area ratio yields a new judgment ratio, used to further screen the compliance of segments. For example, segment N1 has a railing height of 1.1m and a window area ratio of 30%, corresponding to a ratio of 3.67. If the set upper limit for judgment is 4.0, then this segment is compliant. However, segment N2 has a railing height of 1.2m and a window area ratio of only 20%. If the ratio is 6.0, exceeding the judgment benchmark, then the section is non-compliant. The railing parameters and structural position parameters of the selected compliant sections are grouped into a unified set. At the same time, standard field range information is introduced, such as whether the column spacing falls within the range of 3.0 to 3.9m, and whether the extension line length is within the range of 6.0 to 7.0m. Combined with whether the railing height and reflectivity fall within the corresponding standard value range, the records that meet all standard fields are sorted and merged. Finally, the information is archived into three categories: node number, railing parameters, and structural parameters to form a structured parameter summary, which is used as a basis for subsequent construction drawing refinement or parameter adjustment.

[0097] Please see Figure 3 The specific steps of S2 are as follows:

[0098] S201: Obtain the column spacing, eaves extension line, window area ratio, railing height and door opening width parameters from the structured parameter collection table, establish field classification rules based on functional attributes, set threshold conditions related to structural proportion and construction, generate corresponding judgment criteria for fields, and obtain parameter condition judgment standard values.

[0099] To obtain the column spacing, eaves extension line, window area ratio, railing height, and doorway width parameters from the structured parameter collection table, all parameter fields must first undergo unit unification and numerical standardization. Column spacing and eaves extension line should be uniformly represented in meters (m), as should railing height and doorway width. The window area ratio should be retained as a percentage. Then, the parameters should be categorized according to their corresponding functions: column spacing and eaves extension line should be classified as structural attributes, railing height and doorway width as construction attributes, and window area ratio as enclosure attributes. After classification, proportional calculation logic should be set for each attribute field, and typical project data should be extracted for analysis. The ratio of column spacing to eaves extension line should generally be maintained between 0.4 and 0.6 in structural design. For example, if the column spacing is 3.2m and the eaves extension line is 6.4m, the ratio between the two is 0.5. The structural relationship between the railing height and the doorway width is set to a ratio between 0.8 and 1.2, which is more reasonable. If the railing height is 1.15m and the doorway width is 1.0m, the ratio is 1.15, which is also within a reasonable range. The window area ratio is set within a general range according to the building type. For residential buildings, it is recommended to be between 20% and 50%. If the area of ​​a window on an exterior wall is 12m² and the corresponding wall area is 30m², the area ratio is 40%, which is within a reasonable value. Through the above ratio settings and data analysis, threshold conditions related to structure and construction are established for each parameter field, and judgment criteria expressed in the form of ratios or percentages are obtained for subsequent sample record judgment.

[0100] S202: Call the threshold condition in the parameter condition judgment standard value, judge the parameter records in the structured parameter collection table one by one, mark the sample that meets the preset rule for any field, write the marking status into the field information, and obtain the sample quantity triggered by the construction ratio rule.

[0101] Based on the established structural and construction ratio judgment criteria, the field values ​​of each sample record in the structured parameter collection table are judged. Specifically, the column spacing and eaves extension line values ​​are first extracted from each record and their ratios are calculated. If a record has a column spacing of 3.0m and an eaves extension line of 6.0m, the ratio is 0.5, which falls within the reasonable range of 0.4 to 0.6 for the established structural ratio. Next, the values ​​of railing height and doorway width are read. For example, if the railing height is 1.2m and the doorway width is 1.0m, the ratio is 1.2, which also falls within the reasonable construction range of 0.8 to 1.2. Then, the opening... For the window area ratio field value, for example, 35%, determine whether it falls within the reasonable range of 20% to 50%. If it does, mark the field as compliant. Set a flag bit for each field. If it meets the condition, record it as 1; otherwise, record it as 0. After marking the fields in sequence, write the flag values ​​into the corresponding fields of the structured parameter table. After all records have been executed, count the number of samples that were triggered for each construction ratio rule. For example, if 70 records meet the structure ratio, 65 meet the construction ratio, and 45 meet the window area ratio requirement, record the number of triggers for the corresponding field. This information is used for subsequent sample processing and feature dataset construction.

[0102] S203: Trigger the labeling information in the sample size according to the construction ratio rule, organize the labeled samples according to the field structure, remove unlabeled records, and classify the difference type fields into the corresponding sets, unify the sample set format, and generate a surface structure training sample set;

[0103] After completing the field marking operation, the structured parameter aggregation table undergoes sample filtering. First, records with at least one field marked as meeting the criteria are retained, while records with all fields marked as not meeting the criteria are removed. That is, only records where any one construction ratio rule is triggered are retained. The remaining records are then categorized according to their field attribute types. Structural parameters such as column spacing and eaves extension lines are grouped together; construction parameters such as railing height and doorway width are grouped together; and enclosure parameters such as window area ratio are grouped together. The data format and unit representation of field values ​​are standardized across all groups. For example, all railing heights are standardized to the retained... The numerical values ​​in m units are all kept to one decimal place. The window area ratio is uniformly retained as a percentage to one decimal place. The sample fields are assigned standardized field names, such as column spacing named DP, eaves extension line named EL, window area ratio named WR, railing height named HR, and door opening width named WD. The field order is uniformly organized according to the classification order of structure, construction, and enclosure. The sample record number is uniformly named such as N01, N02, etc. Through the above methods, the field classification, labeling and filtering and sample format standardization are completed. Finally, a facade structure training sample set containing the status and standardized values ​​of the labeled fields is constructed for subsequent data modeling and parameter verification.

[0104] Please see Figure 4The specific steps of S3 are as follows:

[0105] S301: Extract window area ratio, material reflectivity and floor height parameters based on the facade structure training sample set, complete parameter positioning according to field order, merge the extracted fields into a unified set, and define classification labels according to parameter attributes to obtain parameter type attribution label values;

[0106] Based on the facade structure training sample set, the parameters of window area ratio, material reflectivity, and floor height were extracted. First, the corresponding parameter fields were identified sequentially from each sample record according to a unified field name format. The window area ratio was uniformly represented by WR, material reflectivity by MR, and floor height by FH. During extraction, the units of the fields were standardized; for example, floor height in cm was converted to m. Material reflectivity was uniformly retained to two decimal places, and the window area ratio was retained as a percentage with one decimal place. After numerical standardization, the three field values ​​were sequentially imported into the parameter set, forming a parameter set with three attributes. Then, the parameters were categorized according to their functional attributes. WR represents the opening feature... The characteristics are classified as open, MR represents material properties and is classified as material, and FH represents dimensional information and is classified as dimensional. Each category is assigned a corresponding label value: open category is labeled A, material category is labeled B, and dimensional category is labeled C. For example, in sample record number N001, WR is 35%, MR is 0.28, and FH is 3.2m. Then its corresponding label set is WR-A, MR-B, and FH-C. During the processing, the field order must be kept consistent with the sample number. Field extraction, unit unification, value normalization, and label generation operations are performed one by one. After performing the above process on all records, a parameter classification label set for each sample can be formed, and each field clearly corresponds to an attribute identifier and type classification.

[0107] S302: Call the classification information of the field in the parameter type attribution mark value, divide the parameters into opening, material and size sets according to attribute characteristics, set the classification numbering rules and complete the mapping, and obtain the parameter classification mapping number value;

[0108] The classification information of the fields in the parameter type attribution marker value is called to classify all fields into three sets: opening type into set O, material type into set M, and size type into set S. Then, the classification numbering mapping rules are set according to the field order in each set. For example, fields in set O are numbered O1, O2, etc. according to their original column order in the sample set; in set M, they are numbered M1, M2, etc.; and in set S, they are numbered S1, S2, etc. This mapping numbering is not based on the actual physical quantity, but on the field's position in the original data and its attribution category. If field WR is in the original... If a field is located in column 3 of the data table and belongs to the "opening" category, its mapping number is O3. If MR is located in column 4 and belongs to the "material" category, its number is M4. If FH is located in column 2 and belongs to the "size" category, its number is S2. This method establishes a mapping relationship between field names and category numbers. All fields are assigned mapping numbers according to their category and original column order. The record numbering results can be organized into a parameter mapping list, for example, WR is O3, MR is M4, and FH is S2. After completing the numbering conversion of all fields, a set of parameter mapping number values ​​containing the category numbering rules can be obtained.

[0109] S303: Based on the association structure between the field numbers contained in the parameter classification mapping number value and the sample number, establish the index relationship between the coded data and the sample record, organize the set field numbers and put them into a unified structure table, and generate the sample parameter card index value;

[0110] Based on the acquired parameter classification mapping number values, a field number aggregation and sample number pairing operation is performed on each sample record. The record number of each sample, such as N001, N002, etc., is read, and the field mapping numbers contained within each record are extracted to form the field number set corresponding to that sample. For example, sample N001 contains field numbers O3, M4, and S2, meaning the associated field set for N001 is O3, M4, and S2. This process continues to iterate through subsequent samples, extracting their respective field number sets to establish a one-to-one correspondence between sample numbers and field numbers. Then, all field numbers are uniformly organized, duplicates are removed, and a total field number set is constructed. Finally, a structured field table is built according to the sample order and field number order. The structure uses the sample number as the row header and the field number as the column header. Each cell is filled with the parameter value or tag value of the corresponding sample under that field number. For example, WR of 35% corresponds to O3, MR of 0.28 corresponds to M4, and FH of 3.2m corresponds to S2. The structured parameter combination for sample N001 is O3-35%, M4-0.28, S2-3.2m, and the structured parameter combination for N002 is O2-30%, M3-0.45, S1-2.8m. After organizing the field structure combination of all samples in this way, the parameter card index value of each sample record can be generated based on this field combination. Each card contains the sample number, the corresponding field number, and the actual value, and has uniqueness and structure attribution information.

[0111] Please see Figure 5 The specific steps of S4 are as follows:

[0112] S401: Based on the encoded data of the training sample parameter card set, call the window area ratio, material reflectivity and layer height parameter fields in the sample, match them with the corresponding fields in the input set, identify the numerical deviation between the fields, and obtain the parameter field difference value;

[0113] Based on the coded data of the training sample parameter card set, the card records are first read one by one, and the field values ​​labeled WR, MR, and FH are extracted. WR represents the window area ratio, expressed as a percentage; MR is the material reflectivity, ranging from 0 to 1; and FH represents the floor height, all in meters. After extraction, these values ​​are matched one by one with the corresponding fields in the input set, ensuring consistency in field names and order. Then, the numerical differences are calculated for the matched fields: the window area ratio difference ΔWR is the difference between the WR value in the sample and the input WR' value; the material reflectivity difference ΔMR is the difference between the MR value in the sample and the input MR'; and the floor height difference ΔFH is the difference between the FH value in the sample and the input FH'. All three difference values ​​are converted to absolute values ​​to reflect the degree of numerical deviation. For example, if the input field has WR of 30% and MR of 0.40... If the FH is 3.0m, and in a certain sample the WR is 36%, MR is 0.32, and FH is 2.7m, then the differences in these three fields are 6%, 0.08, and 0.3, respectively. Before processing, the units of the values ​​should be standardized: percentages should be rounded to one decimal place, reflectivity to two decimal places, and floor height to one decimal place. All field differences need to be normalized in further processing to be compared under the same unit. The normalization process can be based on a linear transformation of the maximum and minimum values ​​in the training samples. The window area ratio should be between 10% and 70%, the material reflectivity between 0.1 and 0.9, and the floor height between 2.5m and 4.2m. Each difference value should be normalized according to these ranges. After completion, a set of field differences containing the WR, MR, and FH difference values ​​corresponding to each sample should be constructed for subsequent deviation measurement analysis.

[0114] S402: Based on the parameter field difference value, identify the overall deviation of the corresponding field of the sample, construct a sorting sequence according to the deviation, mark the corresponding position between the sorting number and the sample number, and obtain the field difference sorting sequence;

[0115] Based on the processed parameter field difference set, the normalized difference values ​​of the WR, MR, and FH fields are extracted for each sample. The overall deviation of the sample is constructed based on these three values. The three difference values ​​are then weighted and summed to form a comprehensive deviation value. The proportion of each difference in the overall value is determined by its weight. For example, the weight of the window area ratio can be set to 0.4, and the weights of material reflectivity and floor height can each be 0.3. In a sample, the normalized difference value of WR is 0.75, the normalized difference value of MR is 0.53, and the normalized difference value of FH is 0.60. The comprehensive deviation value can then be obtained by adding them according to their respective weights. The smaller the comprehensive deviation value, the better. This indicates that the sample is closer to the input set field. Following this rule, after calculating the comprehensive deviation value of all samples, they are sorted from smallest to largest. During the sorting process, the number of each sample is bound to its position in the sort. For example, if a sample number is N012 and its comprehensive deviation value is the smallest among all samples, its sort number is 1. Another sample number is N005 and its deviation value is the second smallest, so its sort number is 2. This process is repeated for all samples to construct a difference sorting list. This list arranges the sample numbers according to the size of the deviation value. Each sample number is appended with its sort number, thus forming a complete field difference sorting sequence.

[0116] S403: Based on the order of sample numbers in the field difference sorting sequence, construct a position mapping table in the index structure, verify the correspondence between the sorting number and the sample card number, and generate the fit sorting index number;

[0117] Based on the order of sample numbers listed in the field difference sorting sequence, a position mapping table of samples in the structure index is established. Each record uses the sample number as an identifier, and its number in the sorting sequence is used as the mapping value. For example, if the sample number is N012 and the corresponding sorting number is 1, then the record in the mapping table is N012, corresponding to position 1. If the sample number is N005 and the sorting number is 2, then the corresponding position is 2. A complete mapping set between sample numbers and sorting numbers for all samples is established. Then, the consistency between the sorting number and the parameter card number is verified. The original number of the parameter card usually has the same identifier as the sample number. For example, if the parameter card for sample N012 is C012, and the sorting number is 1, it should be confirmed whether the first item in the sorting result is C012. If an inconsistency is found, the index position of the card can be rearranged according to the sorting sequence. The newly generated index number is named using the method of C plus the sorting number. For example, if the original card C018 ranks 3rd in the sorting, it will be re-marked as C003. The re-sorting operation of all sample card numbers is completed in this way, and finally a set of fit ranking index numbers is formed. Each index record consists of a sorting position number and a corresponding card number, ensuring that each sample has a clear matching order and index number.

[0118] Please see Figure 6The specific steps of S5 are as follows:

[0119] S501: Based on the path results associated with the index number sorted by fit, extract the column spacing, window area ratio and railing height parameters corresponding to the path, match them with the corresponding fields in the input parameter set, identify the ratio relationship between each field, and obtain the path parameter ratio group;

[0120] Based on the path results associated with the index number sorted by fit, three types of facade parameters corresponding to the path are extracted sequentially: column spacing, window area ratio, and railing height. Column spacing is the horizontal distance between the centers of structural members, in meters (m). The window area ratio represents the proportion of window openings in the entire facade, usually expressed as a percentage. Railing height is the vertical dimension from the ground to the top of the railing, also in meters. During extraction, parameter fields are read one by one for each path number. After reading, the field values ​​in the path are matched one by one with the corresponding fields in the input parameter set to ensure consistency in field type, unit, and data format. The matching process is performed by dividing the path field value by the input field value. Calculate the ratios and record the column spacing ratio, window area ratio, and railing height ratio respectively to obtain the path parameter ratio group. For example, if the input parameters are column spacing 3.6m, window area ratio 40%, and railing height 1.05m, and the path number is P001, the field values ​​are column spacing 4.2m, window area ratio 32%, and railing height 1.20m. The corresponding calculated ratios are column spacing ratio 1.17, window area ratio 0.80, and railing height ratio 1.14. After the calculation is completed, a path parameter ratio group is established with the path number as the index. Each group of records contains the ratio data of the three fields. Each record is saved in the path ratio set to form a complete path parameter ratio group.

[0121] S502: Based on the path parameter ratio group, combined with the column spacing ratio threshold, window area ratio threshold and railing height ratio threshold set in the input parameter set, determine whether the conditions of the path ratio combination are met, and obtain the threshold-satisfied path set;

[0122] Based on the path parameter ratio groups, conditional judgments are performed on the ratio combinations under each path number. This requires combining the column spacing ratio threshold, window area ratio threshold, and railing height ratio threshold set in the input parameter set. The column spacing ratio threshold is set to 0.85 to 1.20, indicating that the path column spacing is allowed to fluctuate between 85% and 120% of the input column spacing. The window area ratio threshold is set to the range of 0.75 to 1.10, and the railing height ratio threshold is 0.90 to 1.15. The path ratio groups are traversed one by one, and the three ratios in each group are compared with their respective threshold ranges. For example, the ratio group for path P001 is 1.17, 0. If the column spacing ratio is 1.14, the window area ratio is within the allowable range, and the railing height ratio also meets the conditions, then the path number meets the threshold requirements. For example, the ratio set of path P004 is 1.23, 0.92, 1.08, where the column spacing ratio exceeds the upper limit of 1.20. Therefore, this path does not meet the conditions and is removed from the path ratio set. In this way, the ratio combination of each path is compared to determine whether all three items are within the set threshold range. If all three conditions are met, the corresponding path number is recorded in the list of compliant paths. Finally, the set of paths that meet the threshold is obtained, which contains all path numbers that meet the three ratio conditions.

[0123] S503: Based on the path number in the path set that meets the threshold, summarize the column spacing, window area ratio and railing height parameter fields under the corresponding path, combine them into the parameter set required for facade configuration, and generate the parameter combination set of the residential entrance facade scheme.

[0124] Based on the threshold values ​​for each path number in the path set, the original path data records are retrieved. The column spacing, window area ratio, and railing height parameter fields associated with the path are extracted again. Each extracted data set is combined in a structured manner to form a facade parameter set. Each record consists of a path number and three parameter items. For example, path P001 corresponds to a column spacing of 4.2m, a window area ratio of 32%, and a railing height of 1.20m, which constitutes a set of facade design data. Similarly, path P006 has a column spacing of 3.3m, a window area ratio of 38%, and a railing height of 1.10m, which corresponds to another set of facade configuration data. All parameter sets matching the path are summarized sequentially and arranged by path number to form a complete facade design input source. Regarding data precision, column spacing is retained to 0.1m, railing height to 0.01m, and window area ratio to one decimal place. The resulting parameter combination set is the residential entrance facade scheme parameter combination set, used in the facade design phase.

[0125] To further improve the controllability of the generated results and the efficiency of model reuse, this invention sets up a semantic refinement and knowledge base expansion mechanism on the basis of S5, and constructs a self-loop process from intelligent generation to semantic optimization.

[0126] In this stage, the system first calls the Segment Anything semantic segmentation model specifically designed for architecture to automatically identify and mask components such as doors, windows, railings, landscape elements, and curtain wall nodes in the generated facade, enabling rapid selection and semantic annotation of local areas. Subsequently, the CLIP model is used to extract feature vectors from the original image and the target modification area, calculate the semantic similarity of light and shadow, color, and material, and adjust the local color ratio and reflection intensity through a dynamic mapping algorithm to ensure consistency between the modified area and the overall facade in terms of light and shadow logic, stylistic features, and material reflection.

[0127] After semantic refinement, the system calls the Real-ESRGAN high-definition restoration algorithm to enhance the texture and increase the resolution of the generated image, making the facade details more accurate and suitable for design presentations and construction demonstrations. The refined image, Prompt descriptors, LoRA model weights, ControlNet parameter configurations, and user interaction data will be synchronized to the enterprise knowledge base in real time, forming a "image-parameter-behavior" ternary dataset.

[0128] The backend system regularly analyzes frequently used and highly compatible parameter combinations, fine-tunes the corresponding LoRA modules, and calibrates the model using data from newly added benchmark projects, further improving the model's adaptability to enterprise design preferences and the consistency of generated results. Through this mechanism, the system forms a closed-loop knowledge evolution process of "creation generation → semantic refinement → data accumulation → model optimization," ensuring that the intelligently generated results are continuously optimized in each iteration.

[0129] Please see Figure 7 A smart interactive terminal for generating residential entrance facades, driven by enterprise knowledge base data and featuring AI multimodal collaboration, including:

[0130] The structural parameter collection module obtains the column spacing, eaves extension line length, window area ratio, railing height and material reflectivity from the residential entrance facade project. The parameters are divided into structural parameter sets, node construction sets and detailed component sets according to their attributes. After unifying the field format, the parameters are merged and organized to generate a structured parameter collection table.

[0131] The structural rule filtering module uses column spacing, eaves extension line length, window area ratio, railing height and door opening width from the structured parameter collection table to set structural proportions and construction threshold conditions. It compares the relationship between parameters and thresholds, identifies sample records that meet any rule, and generates a floor structure training sample set.

[0132] The facade sample construction module calls the window area ratio, material reflectivity and floor height parameters from the facade structure training sample set, divides the three parameters into the opening set, material set and size set respectively, establishes the correspondence between field codes and sample numbers, and organizes them into a training sample parameter card set.

[0133] The parameter difference fitting module uses the field encoding data in the parameter card set of the training samples to call the window area ratio, material reflectivity and layer height in the input parameters, calculates the field difference value and ratio, generates a sorting structure based on the difference value, establishes the index relationship between the sorting result and the sample card, and generates the fit ranking index number.

[0134] The scheme combination generation module calls the column spacing, window area ratio and railing height in the path corresponding to the fit sort index number. Based on the ratio offset relationship between the current input parameters and the path parameters, it determines whether the combination meets the structural combination threshold conditions, filters the path parameter set and generates the residential entrance facade scheme parameter combination set.

[0135] To achieve a dual improvement in efficiency and quality of residential entrance facade design, this invention proposes two core innovations:

[0136] 1. Domain-specific (architectural design) data-driven approach:

[0137] This invention constructs a dataset of benchmark residential projects (10,000+ project images and corresponding parameters), and then uses LoRA technology to perform lightweight fine-tuning on the StableDiffusion basic model. The model learns core features such as entrance scale, proportion, railing design, window ratio, and material reflectivity, generating LoRA weights that can be dynamically loaded for different projects, achieving a balance between high consistency and high diversity.

[0138] 2. Intelligent collaboration throughout the entire process:

[0139] Create a closed-loop workflow system that integrates "inspiration acquisition (multimodal data input) → intelligent generation (domain-specific model-driven) → precise refinement (semantic segmentation and interaction)".

[0140] In the intelligent generation stage, the LoRA model is called in combination with ControlNet constraints and enterprise historical parameter templates to achieve accurate structure and controllable style; in the refinement stage, semantic segmentation + InpaintAnything technology is integrated to automatically identify local components and replace them with high precision.

[0141] The system supports one-click reuse of historical parameters and accumulates the generated data into the enterprise knowledge base, forming an enterprise-level digital design asset that can be continuously iterated and optimized.

[0142] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for data-driven facade generation of a residential entrance portal from an enterprise knowledge base, characterized by, Comprise the following steps: S1: According to the column spacing of the residential entrance facade project extracted from the enterprise knowledge base, the eaves extension line length, the window area ratio, the railing height and the material reflectivity, the uniform field format is classified as a structure parameter set, a node structure set and a detail component set according to the attribute, and a structured parameter collection table is generated; S2: Based on the column spacing, the eaves extension line, the window area ratio, the railing height and the door width in the structured parameter collection table, the structure ratio and the construction threshold condition are set, the samples satisfying any one of the preset rules are marked, and a facade structure training sample set is formed; S3: Based on the window area ratio, the material reflectivity and the layer height parameters extracted from the facade structure training sample set, the parameters are divided into opening, material and size sets according to the parameter type, the correspondence between field coding and sample number is established, and the coding data is collected into a training sample parameter card set; S4: According to the set coding data in the training sample parameter card set, the window area ratio, the material reflectivity and the layer height parameter set are identified, the sorting structure is generated according to the difference value between the sets, and the correspondence with the sample card index is established, and the fitting degree sorting index number is generated; S5: Based on the path result associated with the fitting degree sorting index number, the column spacing, the window area ratio and the railing height parameters in the corresponding path are extracted, the ratio offset judgment is performed with the current input parameters, the path parameter set satisfying the combination threshold condition is taken as the structure combination basis, and a residential entrance facade scheme parameter combination set is generated; The residential entrance facade scheme parameter combination set includes column spacing ratio, window area ratio offset value and railing height ratio; The ratio offset judgment refers to comparing the offset degree between the ratio of two parameters, and determining whether it is close according to the preset offset threshold; The combination threshold condition refers to the maximum offset range of multiple parameter ratios in structure combination, which is used for scheme screening and path fitting judgment; The structure combination basis refers to the matched parameter set referred to when generating the building facade scheme, including proportion, size and material elements.

2. The enterprise knowledge base data-driven precinct portal facade generation method of claim 1, wherein, The structured parameter collection table includes column spacing parameters, eaves extension line length parameters, window area ratio parameters, railing height parameters and material reflectivity parameters, the facade structure training sample set includes structure ratio rule identification, construction threshold condition identification and sample number identification, the training sample parameter card set includes opening parameter set, material parameter set and size parameter set, and the fitting degree sorting index number includes difference sorting structure, parameter difference value and sample card index.

3. The enterprise knowledge base data-driven neighborhood portal facade generation method of claim 1, wherein, The specific steps of S1 are: S101: Obtain the column spacing, eaves extension line length, window area ratio, railing height and railing material reflectivity parameters in the residential entrance facade project, unify the parameter format into standard numerical type, classify the fields according to the attribute dimension, establish the interval judgment rule of the corresponding field, and generate the parameter field specification interval value; S102: According to the column distance in the parameter field specification interval value and the eave extension line length, the structure line position parameter is extracted, the node section number is divided, the railing height and material reflectivity under the corresponding section are summarized, and the corresponding relationship between the railing parameter and the node section is established, so as to obtain the node section railing parameter matching value; S103: The railing height, railing material reflectivity and window area ratio in the node section railing parameter matching value are called, combined and classified according to the section number, the ratio between the railing height and the window area ratio is calculated, the parameter combination is screened according to the upper limit benchmark of the railing window area ratio, and the railing parameter, node position parameter and field specification value are merged and integrated according to three types of attribute sets, so as to generate a structured parameter collection table.

4. The enterprise knowledge base data-driven precinct portal facade generation method of claim 3, wherein, The specific steps of S2 are: S201: Obtain the column distance, eave extension line, window area ratio, railing height and door width parameters in the structured parameter collection table, establish field classification rules according to functional attributes, set threshold conditions related to structure proportion and construction, generate corresponding determination benchmarks for fields, and obtain parameter condition determination standard values; S202: The threshold conditions in the parameter condition determination standard values are called to judge the parameter records in the structured parameter collection table one by one, mark any field sample meeting the preset rules, write the marking state into the field information, and obtain the structure proportion rule trigger sample amount; S203: According to the marking information in the structure proportion rule trigger sample amount, the marked samples are arranged according to the field structure, the unmarked records are removed, the difference type fields are classified into corresponding sets, the sample set format is unified, and the facade structure training sample set is generated.

5. The enterprise knowledge base data-driven precinct portal facade generation method of claim 4, wherein, The specific steps of S3 are: S301: Based on the facade structure training sample set, the window area ratio, material reflectivity and layer height parameters are extracted, the parameter positioning is completed according to the field order, the extracted fields are merged into a unified set, and the parameter type attribution mark value is obtained according to the parameter attribute classification mark; S302: The classification information of the fields in the parameter type attribution mark value is called, the parameters are divided into opening, material and size sets according to the attribute characteristics, the classification number rule is set and the mapping is completed, and the parameter classification mapping number value is obtained; S303: According to the association structure of the field number and sample number contained in the parameter classification mapping number value, the index relationship between the coding data and the sample record is established, the set field number is arranged and classified into a unified structure table, and the sample parameter card index value is generated.

6. The enterprise knowledge base data-driven precinct portal facade generation method of claim 5, wherein, The specific steps of S4 are: S401: According to the set coding data in the training sample parameter card set, the window area ratio, material reflectivity and layer height parameter fields in the sample are called, the corresponding fields in the input set are matched, the numerical deviation between the fields is identified, and the parameter field difference value is obtained; S402: Based on the parameter field difference value, the overall deviation degree of the corresponding field of the sample is identified, the sorting sequence is constructed according to the deviation degree, the corresponding position between the sorting number and the sample number is marked, and the field difference sorting sequence is obtained; S403: According to the order relationship of the sample number in the field difference sequence, a position mapping table in the index structure is constructed, the correspondence between the sorting number and the sample card number is verified, and a fitting degree sorting index number is generated.

7. The enterprise knowledge base data-driven precinct portal facade generation method of claim 6, wherein, The specific steps of S5 are: S501: Based on the path result associated with the fitting degree sorting index number, the column distance, the window area ratio, and the railing height parameters corresponding to the path are extracted, matched with the corresponding fields in the input parameter set, the ratio relationship between each field is identified, and the path parameter ratio group is obtained; S502: According to the path parameter ratio group, in combination with the column distance ratio threshold value, the window area ratio threshold value, and the railing height ratio threshold value set in the input parameter set, the condition satisfaction of the path ratio combination is judged, and the threshold value satisfying path set is obtained; S503: Based on the path number in the threshold value satisfying path set, the column distance, the window area ratio, and the railing height parameter fields under the corresponding path are summarized and combined into the parameter set required by the facade configuration, and a residential entrance facade scheme parameter combination set is generated.

8. The enterprise knowledge base data-driven neighborhood portal facade generation method of claim 1, wherein, The window area ratio is the ratio of the window hole area to the total area of the same section facade in the building facade; The railing height is the vertical height from the tread to the top of the railing when the railing is set at the balcony or corridor, which meets the definition of residential design specification; The material reflectivity is the ratio of reflected light under visible light band of building facade material; The structure ratio threshold condition refers to the ratio interval established between structure parameters; The construction threshold condition refers to the upper and lower limit standard set for the proportion relationship or size relationship between construction nodes; The preset rule refers to the parameter judgment condition set according to the enterprise benchmark project, including the proportion threshold, size range parameter logic rule; The field code refers to the unique identifiable code structure established by each type of parameter in the parameter set through numbering, letter or combination; The difference value between the sets refers to the content difference number between the input parameter set and the code set in the knowledge base, which is represented by set difference set length or proportion; The fitting degree sorting index number is a number sequence established from low to high according to the parameter difference degree.

9. A corporate knowledge base data driven residential entry facade generation kiosk, characterized by, The interactive terminal is used to realize the enterprise knowledge base data driven residential entrance facade generation method of any one of claims 1-8, and the interactive terminal comprises: The structure parameter collection module obtains the column distance, eave extension line length, window area ratio, railing height, and material reflectivity in the residential entrance facade project, divides the parameters into a structure parameter set, a node construction set, and a detail component set according to attributes, merges and arranges after uniform field format, and generates a structured parameter collection table; The structure rule screening module sets the structure ratio and construction threshold condition based on the column distance, eave extension line length, window area ratio, railing height, and door width in the structured parameter collection table, compares the relationship between the parameters and the threshold value, identifies the sample records satisfying any rule, and generates a facade structure training sample set; The facade sample construction module calls the opening area ratio, material reflectivity and layer height parameters in the facade structure training sample set, divides the three parameters into an opening set, a material set and a size set respectively, establishes a corresponding relationship between field coding and sample number, and arranges into a training sample parameter card set; The parameter difference fitting module calls the opening area ratio, material reflectivity and layer height in the input parameters according to the field coding data in the training sample parameter card set, calculates the field difference value and the ratio, generates a sorting structure according to the difference value, establishes an index relationship between the sorting result and the sample card, and generates a fitting degree sorting index number; The scheme combination generation module calls the column distance, opening area ratio and railing height in the path corresponding to the fitting degree sorting index number, determines whether the combination meets the structure combination threshold condition according to the ratio offset relationship between the current input parameter and the path parameter, filters the path parameter set and generates a residential entrance facade scheme parameter combination set.

Images